Got home just after 1800. We've decided since I must be in Prescott for the evening of the 2 Jul, and that Yonkovich must be in class, that we shall drive for eight hours tomorrow. Ick. It's the only way, though. Today was OK...We went out to Buckeye, up almost to Lake Pleasant, over to 19th ave, and back to the school, switching drivers about every hour. Fairly boring...
Chinese food for dinner (Elena really wanted to eat out of the "white boxes"). Before then, I for some reason was heavily snacking on these really tasty wine gums (well, the bag says "English Wine Gums"). Elena brought them from Finland, as did Sara when she was there last summer. They're umbelievably good for such a simple article. They're kind of like "gummy bears" but with a touch of wine added. The bag's almost empty...I've got to find somewhere in the states to get them.... They leave this really nifty tingly sensation for about an hour after you eat them...very pleasing... And, of course, after dinner, I just had to have some peanut butter and chocolate chip granola bars (only Americans could think of such an oxymoronic combination)...
Still working on getting my mail filtered... It would work great...if I'd never touched it, that is. procmail is not human compatible...
I'm kind of in an impatient mood at the moment and am in fact only writing this to keep from screaming at myself (and procmail). I'll shove off early this evening (yes...get up AND go to bed all in the same day!)... I must return to the world of the awake early in order to meet a car 1000 (way to early) for I can be in it for eight hours.... see ya at 1800...
Well, yesterday was just not worth mentioning. Most everything I did somehow involved sleeping, whether it be laying on a bed, sitting on a chair, sitting on the ground, or just plain standing up. Everything else I did somehow involved token ring. BTW, I found out my stupid, stupid problem....I was idiotically assuming that the IBM ISA token cards I had were 16mb...needless to say, they weren't. So, now I have two seperate rings: a 4mb and 16mb. Yikes... I also had to reboot ihpled yesterday. My insesant meddling with token ring drives it insane. Once `ifconfig tr0 down` gets itself stuck in the D state, it causes bind to get stuck with it on it's next refresh. It'd been up for a new record of 13days, 6hrs, and some change.
I finally got tired of waiting for my 6,900 message inbox to open up, so I got mad just gziped and downloaded. It's running through procmail on ihpled right now. I hope it gets done before I have to go -- it's on 20 May at the moment...it's got more than a month left to go.
At 1400, I must report to school...again. Behind-the-wheel is today, tomorrow, and the next. That unfortunatly lasts until 1800. Four hours of straight driving everyday. Not exactly fun, or the best use of my time.
It's made it to 2100 on the 21 May...
This files sure getting large....we'll purge in July...
I'm kind of depressed by the count of usefull days versus useless days, but it would appear I'm going to have to increment the latter... Nothing done today.
Did a little logic mapping of PyxisMail procedures, but after looking at it all, I think I've decided to start over. I mean, rather than have every function access the mailboxes directly, I think I'll do everything through a forking gateway daemon. And have the daemon lock the boxes, keep authentication records, etc. The interfacing between the CGI and the daemon is what's scaring me at the moment. The two logical solutions that come to mind...SysV IPC and FIFOs. FIFO is probably pretty unreasonable since they're only two-ended. That means IPC. IPC is complicated and I don't even know where to begin. I guess shared memory is the best way to go there.
A call from Herbert woke me up this morning. Well, this afternoon is more like it. He wanted to know how to get NT onto his AlphaStation. I just pointed him to the Linux instructions. All he needs to do is convert the flash to ARC instead of SRM. He also gave me a source for a case+power supply for the SPARCstation 2 motherboard he gave me. He says they're US$50. It might be worth it. An SS2 might be nice just to have around.
I did a little documenting on the AIM protocol. Rather, a little correcting. I really shouldn't have coded that much without documenting it. The main problem with protoFAIM at the moment: the packet dump I was basing the code on didn't cover the entire login process! So, I've had to make a new one and just fill in and test my theories about sequence numbers. Aparantly it didn't work, so now the server bails out halfway through phase 3 with data I've never seen before. This may also be my problem, since my program can't send and think and recieve all at the same time like it appears it should. That means...multithreading! Or forking+IPC....
After that, I wrote this. Well, not this...I wrote the above first, then this, then this, then this, then this, then this... Good night.
Not much done today. That is, besides feeling tired and generally terrible, along with getting nothing done.
I was awoken at 0730 this morning and told to come eat breakfast. Considering I didn't get to sleep until 0300, that wasn't too much of an excitement. Then, when I tried to fall back asleep, something/someone always ended up waking me up again for some inane reason. Sheesh...people!
Suppose to leave for Sunset Point at 1030. Sara got a free hour at her voice lesson which was suppose to be over by 1000, but then didn't get out until 1100. She didn't call or anything, so we were all kind of upset that we were just standing around waiting for her.... Finally got everybody out the door 1130... Sunset Point was just as normal as it always is. Windy (a minor side effect of being a big hole in the ground) as usual, though everytime we go, they all seem surpised it's windy... Very boring. Didn't get home till 1530 or so.
At that point, I attempted to sleep, but of course couldn't. So, since the AlphaStation sits right across from my bed, staring me in the eye everytime I laid down, I decided to do something with it. I'll fill you in on why the AS200 didn't have any OS on it.... When I tried to install RH5 on it a few months ago, the install died with sig11's. That disgusted me and scared me all at the same time. And it would seem I never got back to it today. I installed RedHat 5.0/alpha (since I already had the three install disks made, it was not too hard). I did an FTP install from the echinda master (it just happened to be the closest Linux box with an open-faced, easily accessibly SCSI plug). Everything went fine...well, let's say it did everything fine. I had to redo it three times because of my own idiotic mistakes (ie, selecting too many packets, etc). I got it installed. It configured X for me, which was a surprise. It automatically detected my video (a #9GXE64pro) and copied down and configured the right server. Great, I said. After the first reboot, I attempted to start xdm. Big mistake. It got the normal X cross-hatch pattern fine, but when it started to display the RH-ed version of Tux, the entire system froze just after hitting the top of his eyes! Bad news. I just hit the switch and forgot about it. I'll deal with that another day. I've had enough with system freezing for one month...
I decided as tired as I am, I wasn't going to be productive on my proplems with protoFAIM at all, so I decided to start on PyxisMail. That is, my HoTMaIL clone. My idea is to store mailboxes in normal UNIX format so that you don't have to be a dedicated PyxisMail user or dedicated shell/POP user. I'd like to be able to do both. That does appear to be that hard, except for the fact that it means I must use the system for authentication, and not just rely on my own methods. Which seems to mean right now that we'll have to run set-uid root. I don't particularly like the idea of a CGI app running suid root, but it looks like that's the only way to go (unless we have each user use a different copy of the app, which is suid to their username, but that's stupid and too much work. The plan is now to run suid root, get the name and password from them (the same as thier system login and password), and do a setuid() to become owned by thier UID as soon as possible, in order to reduce the risk of serious damage to other users. If we can limit it to that, we can limit the number of breakable (ie, maliciously damagable) parts. It's still not optimal. I've still got a lot more thought to do....such as, how do we keep the user authenticated across their entire session? I'm thinking about using a database, since that's faster and more versatile than me just using a flat-file and flock()ing it. If there's one thing I truely hate, it's flat file databases. Ick. Besides, if I use a database, the system can be distributed by means other than NFS. Not that this system will ever have more than a dozen or so users....
Basically, I got a login HTML page done (actually, I did that last) and a set of authentication and CGI parsing routines. The authentication routines are basic, because the method is basic (nobody ever said you had to be complicated to be secure). The CGI parsing routines are not-so-basic. In fact, they're terrible to look at. I should have done it in perl. But, I like C. I like the challenge of finding memory leaks.... Normally, I'd shoot out a link for you to see the source code at this point. But, I'm not. That's because I've been doing the coding on zeta this time, which is strange for me, but for some reason I did it anyway. No public stuff yet, I guess. As I said the CGI routines are horrid. Actually, they were worse. After staring at them for a while, I scrapped most of them and replaced them with routines that would a tenth as long and 10 times more efficient and understandable (that basically cut the number of lines in the code in half). Sometimes I just think too complicated. Though, the way I was doing a simple string traversal w/ replace operation was quite interesting. Instead of doing a direct traversal, for some reason I was strtok()ing it then combining in back together again. I'd used realloc after each strtok() to resize my temp buffer and memcpy()ed it back onto the end of the temp buffer in the same order they were before, just with every instance of & set to a null. You can obviously see like I did that that's just plain stupid (I replaced that whole mess with 3 lines: a while, an if, and an assignment). But, it did look damn neat when it was coded that oddly. It made it look like I knew what I was doing... Basically, it was the first chunk of code I was proud of because it did something in a really stupid way. Oh, the theme of the evening was replacing as many while()'s as you could with for()'s. It was pretty fun and made the code look neat...and shorter. I was using a lot of whiles + increments, when I could have just been using a for with an integrated increment.
That's about it. Since I didn't get much sleep last night, I'm going to lay down early tonight. The last words read for the evening will probably once again be from the "TMS380 Second-Generation Token Ring User's Guide" from TI. I'm still on chapter 2, but it's probably the most important for me right now. I wish I'd learned token ring before ethernet...then it wouldn't sound quite so complicated! Eeek...I still need to finish "The Linux Kernel" as well...maybe tomorrow we can begin on that again. Eeek...I need to go driving one of these days before Tue... Eeek...I've got behind the wheel on Tues.... Eeek...I said Eeek three times in a row...Eeek! That's six! "Merry Christmas to all, and to all a good night."
Got some code done today on protoFAIM. Actually, I did code for more commands than I have written the spec for, and it's cost me. Now I've got problems and no documentation. Bad idea. I'll have to catch it up later today.
I had to quickly write a little packet comparator in order to figure some stuff out for phase 2. It turns out that all the gibberish we got from phase 1 gets sent back in phase 2. That fixes that, and gets us to phase 3c, which is the last "successful" phase. There's code for phase 3d, but it doesn't work. But, that means there's something wrong with 3c, because 3d is just a read!
Also found out what happens when your seqnum reaches 0xff. Well, you increment your srcid by one and go back to zero. Anybody who's worth their weight can tell from that what I'm going to say... That the srcid and seqnum are actually not two seperate 8bits...it's one 16bit int. That's a discovery, to say the least. And, it makes less sense than having two seperate numbers does. I'm not going to completely redesign protoFAIM to handle this, although I should. I'll just remember it for FAIM and the spec.
Past that, I think just a bunch of packets got put in and code put around them...
That is, between freezes. Yes, they're back...in full force. After one of the freezes, the console came up pulsating everytime there was SCSI activity...I'd interpret that as a BAD thing. And shortly after, fsck failed with sig11. Even after giving it plenty of cool-down time, I tried again, with similar disasterous results. Then, my dad convinced me that this was the same problem he was having with the same model of motherboard. His instance of this board would not run properly function at 200MHz. It would freeze and corrupt data (this is Win95). Well, when I started to think about it, yes, it did sound like a little more than a coincidence. His fix for his board was to knock the speed down to 180MHz. So I did, and it hasn't frozen since. (Why does it always end up he's in some way right????) My question still is...why did it just start NOW!!??? I've had that board with that chip in it at 200MHz for well over a year with no problems until now!
His long-term solution to the problem was just to get another board. That worked...his chip worked fine in another board (a VIA (non-Intel-based) board). Hopefully, I'll get to do the same. I'd really like a dual PPro board. The prices of PPro equipment should be getting real cheap now that there's PII's that have the same features for servers (ie, L2 at full chip speed, etc).
Of course, once the freezes started, the ethernet troubles came back. So I said heck with this and replaced the tulip with a boomerang. Still no-go. I figured it was a little odd for both cards not to work, so I plugged zeta directly into the 10Mb hub in my room. Poof! No more problems... but, it's repeated and it's 10Mb... So, I had my dad try the 100Mb connection on the switch. His machine then stopped liking the network too. Plug it back into the 10Mb port, it works. Well, that obviously points to that port on the switch. That's a BAD thing since that switch has only got two 100Mb ports to begin with, and past that, it's less than 6mo old! No wonder it was half the price of every other switch on the market...
Tomorrow should be pretty bad. Halve to go up to Sunset Point AGAIN. Twice in the same month. Last time was Sara's birthday (or punishment, as she calls it), this time it's Amy's birthday (which would otherwise be the 4 Jul, but she's going to be out of town then). Another all-day waste of time. Oh well, can't be computing all the time... Maybe something might happen tomorrow night...probably night...
Only about 15 more packets till we're done with protoFAIM login! (probably on the buddy lists sooner, though)
Had another short talk with Lanée. Ben didn't show once again. She's still in Diego... She kept complaining about a headache... Also wasted some time watching the US/China presidents' talks on CNN. Nothing too remarkable there. There's one thing I do have to say: those Chinese bands play the Star Spangled Banner a lot more musically and interpretivly than any US band... It and their own anthem were truely beautiful. It must be that socialist commitment...
Another day of meager progress. Most of today was either spent at the airoport, at home attempting to sleep but can't because every five minutes someone wants you to do something for them, or working on FAIM/AIM. As you might have guessed, those are in an order...I'll let you figure it out...
Elena came in fine. The British Air flight from London/Gatwick was delayed fifteen minutes. But that mattered not at all in the grand scheme. Little did me know we were going to have to wait 45min while Customs wasted everyones time! I guess we should have known better.
Elena speaks great english. Very little accent. (I just figured Linus didn't have much of an accent because he'd been here for a while -- it turns out there's not much of one to begin with.) We decided not to go out to dinner because Elena (or anyone else, for that matter) was not very hungry. This heat kind of does that to you...
Oh, I got a Proteon MAU in the mail today. It's not much use to me, but it's fun to look at anyway. It's only got hermophroditic (type 1) connectors, of which I have no cables for. And, all the power supply connections are that wierd standard 19" rack +12v/+5v connector. Hmmm... it's neat looking none the less. Plus, it's neat to be talking about "hermophroditic MAUs" and everything thinks your strange...
FAIM progressed mildly. Most of today was just making some abstract generalizations about the overall protocol. I learned about error detection during Phase 1 of Login, as well as string-length specifiers, etc. Also, protoFAIM now successfully completes the first phase of login. It's up to the point where the next step is opening the next connection. (Yes, you may be thinking..."It did that before! whats up?"... Well, yes, it did do it before. Just not correctly nor successfully. I figured out thats why I didn't get predictable feedback. Then I looked back at the packet data I had hard-coded in there and NONE of the source identifiers matched when they should've and the same with the sequence numbers. It's no wonder nothing worked predictably.) Phase 1 login is now successfully working with predictable results. (Actually, you can change the SN or password fields of the first hard-coded packet and recieve protoAIM will send back an error message with a URL for you to look at!) I've only been able to detect two errors so far, but I think I've found the "general error" byte, so I can detect any error, just not specifically identify it. More to come in that area probably. Phase 2 is where it gets complicated, though. I'll just have to do it one step at a time. I'll hopefully begin moving away from completly hard-coded packets soon too. The mechanisms to store the srcid's and seq_num_origins are already implemented. Brock is hard at work on the password encoding. He says he knows what's going on, so there should be an algorith from him soon. I guess it's a good thing we're actually starting to share code and not just the wild, abstract ideas we normally talk about!
Talked to Lanée tonight. She was on wanting to talk to Ben, but he wasn't on, so she talked to me for a while instead. Aparently, she's having a good time in San Diego. The place she's staying is "right by" the beach. She'll be back Sun...just in time for our driving on Tue...fun...
Maybe we'll do some more work tomorrow...maybe not... Remind me to go onto #corelcomputer on irc.corel.com at 8PM UTC tomorrow. Gotta see what they have to say about NetWinder...
I quickly did another packet while I was waiting... Login Fail (SN/PW mismatch). All these shorter packets have allowed me to confirm my suspisions about EVERY packet's byte 6...packet length. It's the number of bytes from (inclusive) byte 7 until the end. Byte 5 is still a mystery, though -- although it's always either 0x00 or 0x01.
Must go now!
Up early today (good thing? bad thing?). I don't have much time, so we'll just write about yesterday, an exciting thing to say the least.
Didn't get any coding done per se, but did learn alot. I worked on figuring out the proper usage of tcpdump, capturing AIM login sequences under different circumstances, and then attempting to decode the information in the packets. I was pleased with the meager results, but it could have gone better (ie, I could have gotten more done). I seem to have quite a tad of trouble getting down and concentrating these days.
Most of the current work is embodied in the preliminary FAIM/AIM Protocol Specification. First, there's a sample dump of the login process, then the decoding begins. The first packet is completely mapped, with constants where it appears they should be (although I can't verify them with only my two AIM accounts -- I'll have to create some more and make sure they really are constants). I also got awwaiid working on the password endec. All that means is that you may not have to have an ether sniffer to use FAIM... :)
Though the most surprising thing that happened yesterday was what didn't happen: zeta didn't freeze! It'd been up for 13hrs, 44min (!) when I shut it down normally (safely) before I went to bed. Wierd... Amusingly, I guess I could just blame the tulip driver for all my problems and forget about the whole mess...
And every minute I wasn't working on FAIM, I was reading the Yellow Book of Token...some very cool stuff. All I can say is that TR is FAR more complicated than ethernet, even with all ethernet's modern-day additions... Frankly, it amazes me how either of the technologies work at all. How ethernet can possibly get anything productive transmitted on a busy LAN is a mystery (yes, I understand it, I just don't know why it works!). And how token ring can coordinate everything so well without going wacky is also a mystery...how many years did it take someone to design something this complicated!?!?!?
It looks like I'll be getting some EISA boards at little cost (basically shipping). Not SMP or anything, but EISA by itself is good too. 486 (5v, possibly 3v) VL/EISA and the other 486 (5v) ISA/EISA. Both have the "Pentium Overdrive" socket, whatever the heck that is.
As I said, I'll be leaving in a few hours. Have to go to the aeroport with the rest of the family to pick up Elena (Hartikka). She arrives from Finland at 1400 or so (actually, she arrives from London).
Well, this has not been a good day. It was painfully hot and I've had nothing but trouble. Well, ok, there were a few triumphs of the man vs. machine thing today, but that includes some that weren't in my favor.
I now sit here, in the garage, at my dad's computer, running an X server under Win95 that's logged in via xdm running on zeta. Yes...I've given up on the video hangs. But, there've been two things successful here: the SCSI problem and the Tulip problem. I'll explain...
We'll start with SCSI. Ok, so I didn't get a chance to comment in yesterday's entry about the machine troubles after that (that was mainly because zeta wasn't usable enough to write on -- hell, it didn't even boot!!!!). About midnight, it began doing all sorts of wild and crazy things. Mainly X, bash, et al segfaulting, fsck returning sig11, et cetera, et cetera... Well, after all that started I reboot to find that NONE of ANY of my kernels will boot the machine now! When I got up this morning (that was both a mistake and a good fortune, but I was NOT in good cheer!), I dug out my Slack 3.4 CD and started rawriting a few kernels. My goal was to find one that good even complete the partition check on my drives without giving general protection 0000... those things scare me. Well, I didn't have to try too many (one, actually). The aha2x4x.s kernel did the check AND mounted the filesystem AND got the system into X AND WITHOUT ANY troubles at (well, besides all the errors about missing ethernet). That scared me because now I didn't know what could possibly be the problem. Slack3.4 is 2.0.30. I tried my 2.0.30 kernel that I used for almost a year, and it died....why did this one work? Well, frankly I don't know. I still don't know why two versions of the exact same driver don't work the same. So, now that I'd narrowed it down to the AIC7xxx driver, my dad just said: lets get a new one. This was joyous for me because I was tired of dealing with the damn card (that, and I was scared that some of the wierd stories on linux-scsi list about this card might actually happen to me!). So, we found a Buslogic (they're Linux supports anyway) BT-950 (aka, Mylex Flashpoint LW). It does everything the Adaptec did, just at about a third the price. So got that ordered. I was bored (ya, that took a whole 5min), so I started trying again. I began to slowly glide through all the messages that I've collected from the linux-scsi mailing list for the past couple of months. I found a note from a guy who was having the EXACT same problem. He'd gotten no useful replies, so I started investigating his details. He'd gotten it narrowed down to the extended (>1gb) translation (ie, it wasn't turning on, which makes all accesses about 1gb fruitless). A quick glance at drivers/scsi/README.aic7xxx told the answer: force translation on with the boot line "aic7xxx=extended". Well, that worked, and all was well. zeta's now booting again (no floppy disks involved, either). segfaults and sig11's have yet to show their ugly heads (well not really ugly, just not usually very welcome). I'm thinking that at least the sig11's had more to do with the 90degF heat in my room last night, since I don't think it would have anything to do with extended translation. Anyway, the BT-950's on it's way, and I'm debating whether to replace the Adaptec with it or not. I've posted a message linux-scsi asking for advice. The first (and only) message I've gotten back plainly states: "Throw away the Adaptec cards. The drivers are shit, to be frank. The BusLogic cards in my experience are faster anyway, and the drivers are rock solid even with SMP. Even a cheap Advansys card is better than Adaptec (and faster too)." (Jeff Noxon) I've got to believe him, since it looks like many, many othre people agree. That, and I don't think Adaptec makes as good of boards as they used to. Though, if someone wants to tell me WHY this driver just seems to only detect things when it feels like it, please let me know.
After it began booting, the problems weren't over, of course. The ethernet problem was still looming. Well, I think I've fixed that one too. (Though it took a while to fix over the network since the packets were still dropping at a rate of about 25% and I wasn't ABOUT to touch (ie, crash) the console!) I dled the new tulip driver (actually it was brand-new today -- v0.89H 5/23/98) and dropped it into the tree and recompiled. After the reboot, everything seems to be working as it should be. Before this, I was barely able to open an xdm session over the network (and certainly not at a usable speed), and I'm now...well, able to run Netscape (Ahh...finally got to read slashdot.org today) and XEmacs (I'm typing at full rate in it now). Everything seems quite dandy at the moment. Now if only I could use the X server at the console for more than 15minutes without it crashing.... Actually, I think running X across the network is very nice (except for the wierd scrolling/fvwm virtual desktop interactions, and the fact that netscape under this server turns all my penguins blue!). It's actually (truefuly) about the same speed as running it locally. That, and it doesn't crash. I guess that really narrows down where the (hopefuly final) problem is: the video. Maybe I can convince my dad to switch cards with me.... This thing's been running for over an hour now with no problems at... (I'm guessing when I go in and login to the xdm on console, it will die near-immediatly, though.) Hmm...this thing actually handles the GIMP FASTER than locally, at least it would seem... finally (netscape+xemacs+gimp+aim)!=(crash)!!! it's a good day in america...
I'm just glad I'm 2/3rds done with these problems. They've driven me insane. I think I'll actually half to lay down a few hours early tonight just to attempt to relax. Not exactly my idea of a carefree day...
Also had a short late-night chat with Ben an Lanée. Not too much exciting stuff happening. Ben's got to take driver's ed this session and he know's no one...besides that, it's just a bad place to be in general. I'm just glad Lanée was there to talk to when we had to take it.
Well, I'm back on the console. It hasn't frozen yet, but I'm guessing it's just waiting till I try and do some REAL work...of course, I've only been using it for about 20minutes...
Reminder to self: attempt to begin working on a hotmail-style SMTP/POP interface...we'll have to support the "read-anywhere" folks here soon...
Today, the majority was the other way around. I spent most of the day dealing with zeta's freezing. I still have no solution. Although, I do know that once I knocked all plates off the front and shoved a big fan in there, it hasn't crashed yet (that was 2hrs, 43min ago).
I now know the noisy drive is definitly the older Micropolis (the 4110). Sounds like going bearings. The other Microp has been run ragged today doing fsck's over and over (once after every freeze, sometimes twice). Wierd stuff. A few times I'd have to run fsck manually because the automatic one saw detatched inodes. Those were nothing, though -- just some Netscape cache files. One of these instances scared me though: it once (and only once) came up with hardware bad sectors from the drive (my Linux drive). I shut it off and left it there for 20min or so (gave me a chance to eat my one and only meal of the day), and turned it back on. It ran fsck, found detatched inodes, dropped to single user, I ran fsck manually, fixed the inodes, and all was well. No problems since. (It was after this last VERY SCARY incident that I stuck the fan in. Well, the kernel null pointer dereferences from the swapper were also an incetive for the fan.) Oh, and I'm getting some wierd behavior from my 21140 ether...about a 30-40% dropped packet rate -- ie, frames never make it to the cable. So, I'm guessing driver problem. I'll try and upgrade either the kernel or the tulip driver itself. Bad news there too (often I have to reset the adapter every few minutes just to keep the flow going).
Well, it's not too easy to concentrate on much of anything when your machine is crashing every 15min. Just long enough to get the workspace setup and start typing code...and death. Just a waste of time and I stopped trying after a few attempts. Maybe tomorrow. I still have no better understanding of this AIM thing than I did yesterday. [Whoops...crashed again...back.] It's truely amazing how this thing can actually work with completely random nubmers. (I'm still open to suggestions on good ether sniffers for linux! karpski's inability to save packets, and tcpview's inability to show duplex traffic, and tcpdumps...well...is tcpdump.) I give up (for now). Maybe tomorrow. I need some time to relax from all these crashes (they make me nervous to even move the mouse). Well, so I "coded" some HTML for a while. See the new FAIM site and get a free pass to view the code! Well, it's GPLed anyway, even though it doesn't say that. If you're any good at async I/O, let me know. I'd really love to know how SIGIO works, since it sounds like what I need at this point (if not now, soon).
[Attempting to remember what I did yesterday...]
Well, spent a good chunk of the day attempting to figure out AIM's bizzaro "protocol". It's got some wierd stuff which appears to be to be just random garbage that comes up different everytime you connect. Stupid. Made a map for myself of the login process. I made it; I don't understand it. It comes up different sometimes anyway. I'll put it up in HTML form soon, right after I figure it out to a decent extent.
Spent the other half of the day trying to figure out why zeta is being so wacky. I've narrowed a majority of the freezes down to video. I base that hypothesis on the fact they usually happen you, for example, clicking to open up a menu (in Netscape and fvwm, so far). They also happen right after you swith fvwm virtual desktops. That last one can indicate either a video hangup or a VM problem (ie, the apps on the new desktop got swapped out or something). Wierd.
Today's just been a day full of getting near-nothing done on as many things I could find to do, yet not spend more than a little while on each one... Let's start from now and go back...then forward...then back... (there's a reason I go by midend[ian]...)
I've just gotten done cleaning up the web stuff at delphid.ml.org:80. It's not too much different, just added a few things to protect for the future: a place to purge old news items to, added a few for-psychological-reasons-only sentances to the front page, seperated out this log from the root path, and added the bare beginnings of a PST site.
Also worked on the PB160 a little more... It just doesn't like to stay booted: after booting quite cleanly (and IMHO beautifuly) it will just reboot itself at an aparently arbatrary stop after entering userspace. I don't understand! I've posted the problem to the port-mac68k list, hopefully someone has an answer. Also, Michael says that there's no reason why X shouldn't work on the LC520, but that still doesn't mean it's all going to work right. We'll see: I think I'll be going down there to try it out sometime in the coming week.
Did some chatting with awwaiid. Gave him a simple quick-fix for the DNS mistakes I've been making until I can get them fixed in the Monolith databases (all the updates should be pending as I write this). The best fix for this would be to get the pst.com domain and get rid of all these extraneous domains (ie, ironwood.phx.k12.az.us and delphid/ihpled.ml.org). That will be the day... This was all after I'd sat in on SomeNet's discussion/interview with SuperHacker Alan Cox. Some interesting stuff, although not quite as exciting as the RMS one.
Before that, I guess I was attempting to digitally record a tape for Sara so that a little blip of an error could be removed on the interim. It turns out there were some other things she didn't like about that recording, so we just trashed it and went with a better one. No editing needed.
But I guess that story would appear to end with that...except that I nearly cleaned off my hard drive in the process. All the Winbloze apps I found to do digital recording wanted a buffer in RAM the entire sample set. No good in that...no one has enough RAM leftover in win95 for that kind of thing. That made me purse a Linux path... I ended up at MutliTrak 2.2. The problem there: there's no source, only binary, so I couldn't recompile it for 2.1.x OSS. So, I tried to boot back to my old 2.0.33. Big mistake. A mistake which cost me several hours of my day. For some reason booting the old kernel corrupted my filesystem, leaving doubled blocks and all sorts of bizarre things. So that took quite a while to fix (manual fsck seems to take about three times as long as automated fsck). And besides that, I finally ran into the now infamous GUS bug: whenever you hit record, the machine freezes solid. So, I can't record in any way (under 2.0.x or 2.1.x) until that gets fixed (which I think it is under Alan's 2.1.106ac(x) patches). Anyway since it wasn't all that critical, I moved on.
I don't remember quite what I was doing before that... I've only gotten...oh...3.5hrs sleep in the past 24 -- I didn't make it into bed "last night" until about 04:45 (I had to go to sleep at that point because I knew I'd never get to sleep if I let the sun come up any farther -- my room has a west wall and it heats up way to quickly in the mornings). Then I had to get up at 08:00 to eat our normal breakfast that we only have together on Sat's and sometimes Sun's. Needless to say, I never made it back to sleep. I also didn't make it into the shower until about 14:00...oh well. That's what vacations are for... The main reason I stayed up so late: I wanted to wait for NetBSD to finish installing on the PB so that I could erase it all, repartition and start over again, so that it would be done again when I got up. While I was waiting, I continued my reading of The Linux Kernel (specifically, PCI chapter -- I think that was the first chapter where I actually learned something). I never realized how cool PCI really was... I also scanned the NetBSD install docs one last time to make sure the reboots weren't something I was doing wrong... As well as browsing through some of the NetBSD/sun3 and /sparc stuff to see if there were any major advancments I could use on my Suns. I guess I should make it to bed a little earlier tonight so I'm not so jumpy tomorrow as I was today...but even though I never focused on one thing today, I think I got more done today than I have in some time... The problem? I didn't get any of the important, complicated things done. I think I've lost my hacking spirit...someone save me before I become unpassionate and completely bored about this stuff! I think I'll go over to Ebay and make sure I don't owe any more money...though I think I do... MMMmmm...peanut butter granola bars... eek...looks like I owe way too much...damit...
I've thought about opening this up several times today (or rather yesterday), but am only now getting to it (the next day)....
Fairly boring day... My dad finally got his Apple PowerBook 160 in working order. So, I'm installing NetBSD/mac68k on it, or at least attempting to...slowly... It's been going fairly well: I've found a kernel that actually has the PB video support (the 1.3.2 Generic one doesn't), and the ADB works. I've got all the sets downloaded and mkfs'ed the partitions (although I didn't remember to make a swap until AFTER I'd started the mkfs and install), and started Installer, which appears it will take all night. I don't know what I'm going to do about swap. The PB160's 10mb of RAM may be enough to contain it for a little while, but I doubt it. I may have to do this all over again... And this is only on a 400mb external drive...he wants it on the 80mb internal...eek...I don't think a functional system can fit in that small of a drive! Also, this thing doesn't have any network interfaces, so I've had to use a couple different levels of sneakernet (average sneakernet: floppy disks; advanced sneakernet: large SCSI hard drives). I wasn't about to use floppies to transfer 30megs of set files... Although I wish it had ethernet!!! I really wanted to telnet into a PowerBook! :) You know, just for the fun of it! Hmm...maybe after it's all running, we can do a cute little SLIP setup with slattach...that should work...now where's that serial cable...
Of course to accomplish all the above, I've had to rip apart a couple of my other Macs. Fortunetly, the damage is not permanent. Having only one Mac floppy drive around is kind of a pain. But, the big II can't handle HD drives anyway (no, I don't have the new ROMs), so that's the reason the IIci is now on the network again.
Before all this Mac stuff started, it looked like it might be a pretty productive day. I guess not. zeta's hung up several times today (yesterday), but I still can't find the cause. At least it happens less frequently than it did with 2.0.33. I've changed the mem=128M to mem=127M as per the advice to someone else on the lists. I thought that might be making the transition from real memory to swap a little difficult, but I have no proof that that is what has kept if from hanging in the past three hours since I did it (three hours was the point at which I rebooted from the freeze then). I've opened many many things all at the same time to try and force it into swap, but only with minimal success. The farthest I got into swap was 304kb, and even then it still had 7mb of real left! I think I've finally found the perfect amount of RAM for this machine! The other thing it could be is the GUS driver, which Alan says always has problems.
On to todays token news... After announcing to linux-net that anyone who could write a driver for these SMC EISA cards could have one, it was promptly noted that there was already someone started on it (and others). Or rather, an entire project with much more interesting possibilities than mere token ring drivers: Linux-SNA! The goal is a complete IBM SNA layer for Linux. Wonderful! One more interoperability plus (now Linux can talk to all the S/390's and AS/400's out there...) Of course, as part of that, a fully abstracted token ring interface with support for at least a dozen different cards, along with channel interfaces, etc. Great stuff. It still may be a while before the SMC's supported, but at least there's a plan. (And hopefully it will knock a few of the bugs out of the current token ring driver.)
That's it for today, about yesterday anyway...
Brock's back, in a somewhat mysterious way... A quote from his first and only message to me: "What the heck happened at your house? I login to find a strange computer who doesn't seem to know me! ahh!". And with that, the message was done. Anyway: summer school is over. Finally. Just about the longest 14days of my life right there. (But, at least I got plenty of sleep.)
I just picked up a BusLogic BT-742S EISA SCSI card from a guy in Canada: $10.50. I love this: you get great cards for dirt cheap because no ones got an EISA machine anymore! The place I got the BT-747s for ihpled nearly begged me to take it because it'd caused them so much trouble. I haven't had any problems yet. I think I'll swap the 747 in ihpled with this 742, since ihpled only uses it for tape, which is alot slower anyway. There's also a Gigabyte EISA/PCI Pentium board with P66 going for relativly cheap, but I don't know. I'd like it...but I don't have the money now. Hmm...there's always parents, I guess... Anybody want to loan me $50?
I got the TI "Yellow Book of Token" today...hehe. TI's books remind me a lot of National Semi's. Not only do they explain their implementation, but they expain the specifications behind it as well. For example: NatSemi's "Ethernet Handbook". One of the greatest books on ethernet I've ever seen (and, best of all, it's free). This TI token ring book has the same (for example, chapter two (about a third of the book) is dedicated to the explanation of the IEEE 802.2 (LLC) and 802.5 (TR) standards at an abstract, platform-independant level). Great stuff. Also, TI sent me the book for the exact chipset I needed (the Second generation TMS380x(x)), even though it's probably out of print -- the year stamped on the cover is 1990 (yes, 8 years ago). The make it look so simple... hehe Now I just need to finish reading The Linux Kernel and I can begin... ya right... Then get the info for the SMC tokenmaster32 cards and we can have a 16mb ring with 15.6mb throughput...hehe
Jared (PUSD) got one step further to running Linux on his Mac. My dad stripped a 68882 from a IIci today (I guess he was bored) and plugged it in. An LC520+FPU, that should work. If Linux doesn't work, at least NetBSD/mac68k will. This will make two Linux boxes that infultrated the PUSD walls...hmmm...NetWare3+NT4+RH5+Deb...that sounds pretty lethal...I'm betting for the Linuxies though. I guess I'll work on that sometime next week... "Life Goes On"...
What's that? You mean awwaiid finnally came out of silence??!? I'm shocked and appauled! (You know, both at the same time!) Yes, that's right... just long enough to ask for his login name to be changed and personal cgi-bin to be setup...
Sheesh. Well, I finally got tired of dealing with eagle. I've moved all the critical stuff over to ihpled (specifically, my mail and the main web site). I think it will work alot better if we don't have to worry about eagle's disk space problems or the NFS crashes that've been going on. Slack 3.4 came with procmail pre-installed, so the mail was a clean move (just a matter of FTPing everything over). The www-80 was the same. The working copy of the web site is now /usr/www/80 on ihpled (as opposed to /usr/www/htdocs-8002 on eagle). I've left my home page on eagle (sym linked it), but moved everyone elses. I've also rid myself of all .forward files. This should all be in order for a clean transition to our own domain. Soon, I hope. I've also setup the personal anon FTP directories (~/public_ftp->~ftp/users/$login). Hopefully this will be the last of the AmazingCrashingInternetServices(tm), for now at least.
The fun with token ring continues. The SMC TokenMaster32 EISA busmaster token cards arrived today (four of them). [Note: if you feel you can write a decent linux driver for these things, let me know, I'll send you one!] As far as getting a functional ring going, I've made it as far as two lobes, then it's over. I can have fogbow and echidna master on the same ring and can ping across fine. BUT, when I turn on one of the echidna nodes, the entire ring dies and crashes hard. Once the node's kernel hits "tr0: Initial interrupt...", the ring crashes. Many strange errors on all the lobes. I can't understand them. I also can't get the ring to recover after I shut down the afflicting node. It's not always the PS/2's that cause the first problems either. Remember before? I couldn't get fogbow or the echidna master to cooperate on the same ring as the 3Com in ihpled? I haven't had the guts to try that one again. I've tried a couple of MAUs and the same thing happens. I tried...
Hehe...with all these bizarre token ring cards, maybe I could become the Don Becker of token ring... I really need a EISA motherboard over here. I wish I could knock fogbow completely apart again and rebuild it from scratch as an EISA system. Anyone? 486 EISA motherboards?? Surely someone has got one!?!?!
Another Promise card arrived today. So, I decided to resell it on EBay. Maybe I can at least get my money back for the card I don't want. Hopefully...
It appears that Brock made it back. He flashed on AIM, but disappeared before I could talk to him (that was probably on purpose). Maybe tomorrow. Which is interesting...the last day of summer school.
Jared from PUSD mailed me today, wanting to know how to get Linux onto the MacLC520 he's now required to stare at. I'm sure it's just to spite the person that made him take it... hehe That seems to be it for now...errrr
Yet another no-thrills day...
I've been attempting to help Hawke out with using his 3c905B (for his DSL connection -- wish I had one) and Linux. It's odd because he's using the right kernels, yet the card still isn't detected properly. So, I was going to compile a kernel for him to try out, but something has broken on my machine! When I attempt to compile a 2.0.33 kernel, I now get bunches of wierd errors: undefined symbols (when they aren't really undefined), missing entry points, non-gcc compatible headers, etc. I don't know what happened. It makes it to compressing the image and fails with undefined symbols in arch/i386/compressed/misc.c. So I thought I'd just use the vmlinux image straight (uncompressed). Wrong. It's severely broken as well. I'll just compile a 2.1.105 for him. He had to remove from his system for now anyway (I'' just have him rawrite and boot a raw floppy vmlinuz).
As stated elsewhere, ihpled started being a little funky today (and somewhat unmanagable), so a reboot was in order. See it's maintence log for more details there. As I said there, it's amazing how much you take DNS server for granted... I had machines crashing and hanging all over the place. eagle also died today. The reason: nfs. Why I'm having so much trouble with NFS in the past few days, I don't know. The 2.1.103 on ihpled has always given some strange NFS errors on boot, but they seemed harmless once the system was booted and everything worked fine. eagle today got stuck in some sort of look while spitting out messages like "RPC: cannot transmit". Someone want to explain that message to me??? I rebooted and all seemed fine, until it tryed to start the rpc portmapper, at which point, it began issuing those same messages once again. I messed around in the NetBSD kernel debugger a bit, trying to kill portmap, but with no success. I just hit c and then hit ctrl-c and it continued booting without further problems. All of eagles and ihpled's NFS services appear to be functioning fine at this point. The result of all todays crashings: all my mailing lists have been unsubscribed because of too many bounces. booo...
I guess today wasn't quite as thrill-less as I thought...
I guess I'll be mailing Alan Cox and see if I can get prelim source for his LanStreamer token driver. TI is sending me the book for the TMS380C5x chipset that's on one of the token ring cards I've got. It's a full DMA board, so it should have reasonable performance. Now I just need to finish learning how to write device drivers, how DMA is suppose to work, etc etc etc... I read a little bit of The Linux Kernel everynight before I go to bed... :) Someday, I'll be able to do this. Hopefully, though, Alan has abstracted the token interfaces enough to make it easy to write. From what it looks like at this point, since there's only one token driver total (ibmtr), there's not much abstraction at all. That will have to change...
I've also begun "working" on the cluster again. I've dug out the two machines I've already started on, along with a third one I quickly made like the other two (remove lid, unplug ESDI, crack back panel blank, add token card, replace lid, mess with stupid crooked screws, mess with them somemore, get really mad, stop before they're all the way in, etc.). Stack1 is way too tall at this point (it now has a total of 10 machines, that's about 2.2m high, 7 machines on the top segment alone). This time, though, I've put some foam padding between avix and the echidna nodes. I don't like having scratched up vaxen... Of coarse, the unstable padding makes the stack all the more volatile... hehe
That's it for now, I think. Only two more days left of getting a 5hr nap every morning from 7:30 to 12:30... oh well. Now I'll just have to stay up till 4:00 and sleep till 14:00...you know, like I usually do... Oh, Ben came home today. Brock should be home tomorrow, I think. Elena arrives from Finlandia on the 25th of June (ie, in a week).
Days like these are about as average as my life gets. Very boring...
Finished up another worthless day of summer school -- now only 3 days left. Once again, nothing done for the first six hours of my waking day. Of course, that was pretty normal for today: I didn't get anything else done today either. This is by far the most boring day I've had in some time...
It looks like someone else has volunteered to write the Chips&Tech 82C9001 chipset capture driver for Video4L. (This is for the original ISA-based Video Blasters, which aren't really that bad, but they were just kind of slow if you can't use the overlay -- Good for TV, bad for capture. Besides that, the captured pictures had somesort of unremovable green tinge.) See here for more info about that (and related) card. The person who volunteered for the driver is Josep Torne.
My meager attempt at compiling Gnome and company seems to be failing in every direction. mico won't compile, complaining about missing headers. The problem now: No space left on device. I've also had to mess with the configure script that comes with gnome-libs-0.20 because it doesn't seem to detect imlib. I don't know why. It was easier to avoid the problem than to fix it. Ok, now manually adding -lpng to all the makefiles.... that fixed that...done. That's one gnome-*-0.20 package down, 8 more to go... Oops. And we end there for tonight. It now won't comtinue without mico, so I'll stop. Zzzzzz....
I figured I'd better write something before today turned tomorrow and yesterday turned the day before...
Finally got a real OS on the new from-parts machine (the one that I put together because I couldn't stand having delphi's old case left empty). Slackware 3.4 install went fine, with a few exceptions. The IBM IDE harddrive I'm using to boot that machine (I can't boot off the SCSI, because it's a dinky Advandsys BIOS-less board) is wierd. I can't get LILO to work with it. So, I just put loadlin in autoexec.bat on a DOS partition... Not a good solution, but it works. Everything else went okay. It's name: fogbow. For those who don't know, a fogbow is the same a rainbow, but occurs in fog instead of pseudo-clouds. Fogbows have no real relation to rainbows as far as creation process goes, just in the reason they happen: reflectance from small water vapor droplets. Fogbows are more difficult to view than rainbows because fog is denser than the humid air that rainbows normally occur in. Now you know about fogbows.
In my search for a good Eudora replacement for Linux, I've come to move my side-grade to Gnome from SomeDay to today (or tomorrow anyway). I've started downloading everything I think I need. I like the Balsa mail client. I also like everything else about Gnome. This will probably help convince me to try Enlightement again.
I finally spent the 5minutes it took to plug the other half the VCR into my machine (more specifically, the GUS). I now watch video and audio, together, from the same side of the room. (Before, I was watching the telly in a window and listening from the real telly thats located to the rear (out of my vision, but withing my hearing range).) Now I have four jumper cables hanging out of the backside of my machine (don't touch anything!). None of them have fallen off yet though. I also found that xawtv allows you to tell it where to put it's volume control, so I can now control the volume just as if it were plugged directly into the video card (ie, using +/-/enter). Alan Cox confirmed the problems with the Bt848 getting it's sync thrown off way easier than it should. He says his Happauge WinTV does it too. Alan also noted that the older drivers recover from off-syncing better than these newer drivers. Oh, well. Someday, someone will fix it (maybe even me). For now, I just change the input to the other port while I'm changing channels, and turn it back when I'm done. (Yes, just the minor sync flux between channels throws it off.) The Panasonic VCR in my room is making me angry once again: it refuses to tune CNN (36), Nick (37), or Headline/CNN (38). Just snow and it won't let me fine-tune them. Also, it seems that it's channel 2 has somehow gotten tuned to the real channel 5. But I'm not complaining there -- I think QVC (the normal ch2) is repulsive regardless.
I think I'll quit for now. I must get up for school soon. Only 4 days left...
Been a pretty good day so far.... Got up at 8:00, ate some, transfered Sara's AIM config from her old comp to her new one (I wish someone knew her password so I didn't have to transfer it via the registry! I still need to add the ESI printing stuff too...), then went back to bed and slept till 12:30... Then, of course, I got up again, as it was getting too hot in here (I can't sleep when it's more than 85degF!). I then continued messing with the Diamond video card...
So as of about 14:30 today, I now have a working DVC1000 in zeta. It works. Actually, all of the problems I had were in software (ie, the v4l drivers). The main problem being that the video address was not getting passed correctly. I fixed by recompiling the kernel without the drivers in, but as modules, then manually loading the BTTV driver with a forced buffer address (insmod bttv videomem=0xfd4). One point of confusion was the address specification: they only want the first three digits, not the zeros. I guess if I read the docs more carefully I would have understood that better (ie, it says specifically that you should specify the address shifted 20). Both tvset and xawtv are working fine. I think I may have a problem, though: after about fifteen minutes, it looses the video sync for no apparent reason. That was coming off the Speed tape; I'm trying broadcast television now. Hopefully, it was just a minor tracking error that slowly messes up all the Bt848's UltraLock stuff. I also still can't freeze/capture images. Just the overlay works for now. xawtv comes back with some 'no buffer' error. Wierd, but at least it's a software problem (I'm getting tired of pulling this machine in and out!). Also, no sound yet. I need to grab another jumper cable and wire that up, but that can do later. But both of those absenses are worth it. This video is great. Especially for an el cheapo card. I'll be writing up some stuff and sending the info back to the guy who sold it to me, so he can give it to the rest of the people he sells this aparently useless cards too. Also, in xawtv, you must set the input to Composite1, not the default Television. I should hack on the driver a little so it doesn't even have that problem (since this cards only got two inputs, there's no reason to have four options (ie, two extra: tuner and Composite2). Maybe later...first I must find a better way to connect my VCR than two jumper cables and a resistor lead... None of the syncing problems when coming off the cable yet (this is a good thing)...
It's now an hour later, and I've seen no sync errors coming off the VCR tuner. I'm guessing the VCR I was using was having a little trouble keeping a steady sync for some reason. Oh well. Still no sound or capture yet.
I'm now writing of the 12th on the 13th. That's pretty much methaphoric to today. Useless. That is, useless and late.
I got one of those Diamond DVC1000 video capture boards today. Looks just like the picture. It'b basically a Bt848-on-a-stick. There's not much else there. Now I know why they were so cheap: they've only got SVideo inputs. At least that's what it looks like. But after hard examination of all the available details I could find... I know one is SVideo (though with a few bits of of other things mixed in) and the other is a proprietary digital camera interface. The SVideo input, though, seems usable, though I have no SVideo equipment (yes, there's some articles of archania that we don't keep around the house). But, you probably have sensed that that's no deterent. After carefully studying the Bt848 databook (all 145 pages of it), I think I can use the SVideo plug and just feed a composite signal on the Luminance pin and short the rest to ground. That's the plan, anyway. (I'll ask my dad about it when he gets up in the morning -- he knows far more about this than I.) Since the Bt848 can take 3 composite inputs or 2 composites and an SVideo, there's definitly hope. If the above doesn't work, I'm sure there's a way of patching directly into one of the other MUX inputs and add a composite. The thing I'm still confused about, though, is where the camera input goes into the Bt848. It appears to be SVideo as well, but that's impossible, since there can only be one SVideo input. This might require further research. That would limit that down to being straight composite, which means there's no need for trechery on the SVideo port -- we can just run it directly into the camera port. Going to garage... [tap-tap-tap-bang!whoosh...]
Ok, I'm back. I just spent an hour and a half (almost exactly) tracing PLCC leads on an 8-layer PCB. (Fun, eh? Actually, it was.) I've learned plenty. It does appear that the camera port is composite (and is connected to the MUX1 lead of the Bt848). The SVideo port is SVideo, of course, and is connected to the MUX0 lead (I only checked the luminance (Y) pin, so I'm assuming the rest is wired correctly as per the Bt databook). MUX2 is shorted to ground, which disables it (too bad... I wanted all three inputs!). I've gathered near-complete pinouts of both the ports. The camera port is mostly grounds and +5v's, along with the one composite, and I've only not figured out what one of the pins is. OOOOO!!! I just remembered what that could be: audio in! I think those cameras had a little microphone in them. I'm guessing thats what it is (I didn't trace any of the audio side of the board, so it's possible). So, it looks like the best bet is to run the composite through on MUX1 via the camera port (or severely hack up the PCB to turn on the currently disabled MUX2, but that's more work than I'd like to do). BTW, the extra pins of the SVideo port (there's actually only four pins in the SVideo connector spec), are two seperate +5v and a +12v (that makes seven conductors total). I don't know why they need those. Firstly, I checked to make sure the SVideo Y-pin didn't go directly into the YIN lead, since that would disable the MUX compeletly. Much to my satisfaction, MUXOUT is correctly wired to YIN and SYNCDET, meaning the Bt Mux is wired for at least two devices (which it turned out later, that it was). Ok, that's enough for now. Sometime, I'll fully document those connectors in my usual style (ie, writing up a web page and making a big deal out of it). Since it's nearly 2am, I should probably be doing something else for a while. This was a nice relaxing short-project, though... Hehe...just a short idea...extend the Bt's I2C bus to an external tuner...(it's bus is empty at the moment). Also, I booted up a 2.1.105 kernel with V4L compiled in. It detected the card just fine, but it called it a MIRO. Oh well.
On the non-computer side...Sandy left back to Switzerland today. Aparently, with much high-action drama: families running from terminal to terminal and back and forth from gate A1 to A10...hehe. She finally left, but made it to LA too soon (she had to take an earlier flight because the one she was scheduled for was going to be delayed, which would have made her miss her connecting flight (Luftounsa) to Swiss country). Lanée was leaving for Tucson today, will be back in time for school on Mon. Ben will be back Tue. Brock will be back Wed. Hmmm... I think that's everybody. BTW, I didn't get to see Sandy off because I had to go to "school"...grrr... Anyway, I think Elina will be coming here in a week or so from Finlandia. I believe shes staying a month or so. Then she leaves...then Amy leaves for NAU...then Sara (well, all of us) leave to Lindsborg, KS...we come back (without Sara)...then I must begin school (ick). And this whole mess starts over again... I'm rambling on about things that have nothing to do with today. I should stop. I know there was something else I wanted to write here, but I can't remember. I'm still looking for EISA and/or PCI 486 mobos, along with dual and quad 486 mobos (preferably, a EISA/PCI quad-486, but I think that's pushing it). I shall be writing again later in the day, I suppose...
Oh, I remember now. I finally took the plunge and upgrade afritz/zeta to 2.1.105. If everyone remembers, the last time I tried a 2.1 (.76 at that time) kernel on this beast, it died hard right after mounting the file system (although I'm pretty sure I could have fixed that by shutting tagged-queuing off, but since I really did need 2.1 kernels at that point I didn't care). You see, now I need them. The frequency at which 2.0.33 was freezing my machine up cold was increasing way too quickly (it didn't three times in the same hour last night, and twice in an hour today). I thought it was the video card (or rather the Xserver for the video card, since it's beta code), but I swapped video cards with my dad's machine (an S3 ViRGE/DX), and it did the same thing. So I decided this can't be a hardware problem, and so far, that's held true. It's been up for almost 5hours now under heavy graphics usage (with the I128s2 back), and it hasn't crashed. Of coarse, it's been using 2.1.105, not 2.0.33. I guess the problems of that kernel were finally catching up to me. Oh well, at least it's fixed. The upgrade to 2.1 went painlessly. I only had to upgrade LILO (I was still using v16, I put in v20 -- I needed bzImage support), net-tools (upgraded to 1.45), and modutils (uped to 2.1.87). I think that's all I did. 2.1.105 booted cleanly with no problems with the tagged-queuing this time -- apparently, the driver now turns it on on a drive-by-drive basis. Of course, since both of my drives are Micropolae, they both have broken tagged queuing, and the driver properly didn't turned it on. I've also noticed a 10% easy performance increase. Probably because of the new VM code... I love it. Works great. I don't know why I didn't try it earlier. (There was more than the freezing that made me upgrade: Alan's Video4Linux was getting quite enticing, and I need some way to drive that Diamond, right?) I think I remembered everything this time...
I knew announcing this as a 'near-daily' log would come in handy eventually...
I can now legally be on the road (though not by myself, yet -- that's not until a few weeks from now). Woohoo... (sarcasm should be noted...) I can say it took longer to get to a DMV that was open that it did to get the permit once we got there. (The one in Sun City, is closed, btw -- just as I'd been told yet no one believed me.) We ended up out in Suprise (at about Bell rd and Dysart). Waited for a little over an hour, compared to the normal waiting time at the other locations of anywhere between two and four hours. It would have been shorted if we hadn't have had to wait for their AS/400 to come back up (of course, it went back down right before I was done and I had to stand there for 15min while it came back up...for a second time!). I'm guessing it was USWest's wonderful leased-line service once again ruining my day... (How do they always know where I am????)
There's been a new house record for uptime. Set by eagle, the new time is 95days, 7hrs, 37minutes. It was finally starting to go wacky and I had to reboot it (luckily, I caught it while I could still do it on my terms -- it looked like it was about to crash itself). After the df command stopped working two days ago, and samba from the winbloze boxen began to crash win95 yesterday (yea!), there was little hope of saving it. I don't know what the problem was. I know when I rebooted I got several RPC errors that were kind of odd. (df and smbd would just hang, showing a status of D in ps. In that state, there's no hope of killing them, even with SIGKILL.) I don't understand. But it works...df and all... ihpled is also still happily multi-processing away... current uptime is 2days, 20hrs, 28mins. That's the longest it's been running yet (would have been longer if it weren't for the token ring incident).
Something more happened yesterday... AZPE finally called me back (yes, after three months). They had a good excuse (though it probably shouldn't be listed here). Nice people. I just sent them a sizable bill for my services along with a recommendation to use the hosting services of Llama Communications. For those of you who don't know, they're a Linux/Apache-based web hosting company with average prices and a BeOS-style web site.
After that late hard night/morning, everything appears to be working functionally (including me, luckily -- at 3.5 hours sleep, I wasn't sure how long I'd make it).
Well, that was a long break. It's now 19:44. That new Samsung/Compaq subsidiary is quite a good thing. Now if Compaq hadn't layed off 15,000 Digital people, this Digital-Compaq thing would be a good deal. [Moment of silence for the lost Digital Equipment Corp. It was great while it lasted (all 25years of it).]
I think I'm going to upgrade zeta to 2.1.103 (since I know it works). It's currently at 2.0.33, which works fine, it just doesn't have all the cool new features that the 2.1.xxx's got (ie, the uber cool Video4Linux). I think I'll run a backup of the machine through ihpled first (I need to test that functionality anyway). What's the best way to do network backups: rmt or nfs? That be the question of the day... I'll just do a short run on the DAT since I know it works (I can't seem to get the Travan to work properly with tar -- it rewinds itself after every set! grrr...). Hehe...I just read up on the subject a little more... Shows you how much I backup... After staring at the gzip manpage (of all places) I remembered that you have to use a seperate device name to keep it from rewinding! zeta is now being beautifully backed up onto the 4gb T4000 Travan. It streams a lot better under Linux than it ever did on Windows, btw. (Though thats no big surprise.) You must use /dev/nst0 instead of just straight /dev/st0. Duh! Anyway... The network is quite active right now (lights are flickering and flashing at an unbelievable rate all over the place!) I'm also copying up a rip of the Office95Pro CDs to the ihpled RAID....
No go on the DMV today. The line was out the door (no sarcasm -- I'm guess there was about four hours of wait there -- we left as we couldn't even park). The new rumor is the Sun City DMV office didn't close after all. I'm guessing there will be a trip to there soon.
As I've said the Promise came today. That was great. Finally, that credit card mess seems to be over. But, something tells me it isn't...
Sara's new machine has gotten far enough to be given a DNS entry (sara\tIN\tA\t198.60.150.206). There's some pretty bizare problems going on with the ether card. (A note: if you ever want to use ether with win95, be sure it's installed while you're installing win95 -- it's a pain in the ass, and possibly even impossible, to install later). I think he's just going to reformat the thing anyway. It would also seem there's a wierd option in that VIA mobo's BIOS about EIDE timing. With that WD drive, it seems to need to be set at "normal" and not "enhanced". Otherwise, the drive will work, it will just not all the time (ie, it will be working fine, then just stop and sit for about half a minute, then start again and continue for a while, then stops, etc..). "Live and Learn"...
Also, the IBM token cards Russ Blakeman sent me (thanks, Russ!) came today. Two of them appear to be the same card, but the third is somewhat longer and doesn't get recognized by Linux at all. They're all dated 1988, btw, so they're classic by modern definition. The two shorter ones get recognized, but they don't seem to work with Linux. The adapter turns on, it's starts passing tokens, but whenever you attempt to send something across it, it starts giving some wierd errors, and then starts flipping on and off the ring. I'm guessing it's an interrupt conflict problem (I've actually only tried one of the cards). Since I don't know how to set the 12 DIP switches that are on any of the cards, I'm stuck for now. We'll see... (this should be fun, right?) Right now, the echidna master node (where I was trying out the cards) is sitting out there flipping the ring on and off (the relays in the TC4050 MAU do get annoying after a while according to my father). I can't down tr0. It just sits there. I can't rmmod ibmtr.o because the if isn't down. I'll have to kill it...
Ooo...almost forgot. I'll be driving for "Behind-The-Wheel" on the 30 Jun, 1 Jul, and the 2 Jul. From 1400 to 1800 each day. This will be with Mr. Yonkovich from Ironwood HS, someone I've never even seen until today. That pushs Lanée's SD vacation to the 3 Jul to the 17 Jul, two weeks (we wasted a way-too-short amount of time in class this morning drawing a calendar so she could decide this -- she claims she always carries a calendar with her -- that will teach her to forget). It wasn't optimal for her, but at least now she has a schedule, which seemed amusing to her (ie, like it was a first time...). We're half done with summer school... only 7 school-days left until it's over (9 days, otherwise).
Eeek... It appears the Stack in the middle of my room fell over in the middle of last night... I came home today and it still hadn't cleaned itself up yet. Maybe tomorrow... In the mean time... well, you know (include memoirs of dodging flying PCWeeks, CRNs, and Network Worlds, along with the joyfull and jubilent sounds of a few MountainDewTowerOfAluminiumCans finding a new resting place in a somewhat heterogeneous way a few feet beside their former location..."Adam!! What's going on out there????"...). BTW, did you know stuff happened today? (If not, read slashdot.org, so you can keep up to date.) Down with Windows/Intel/Wintel/BillG!) ["Last message marked for deletetion."] Tomorrow gives us another yesterday, so make today nice, so you can have a good tomorrow! (ok, so I felt like making a corny inspirational quote to end the evening!) Ahh....SYL...
ihpled was about the only thing that got done today. I stared at the GNOME site while the hda of delphi was copying to the md0 of ihpled (10mb/sec ether is slower than I'd remembered). But, it gave me some time to look at what's going on over there. Great stuff! I'm thinking strongly about trying it out One Of These Days (tm). Overall, everything went smoothly. (zeta was happily "deffering" my incoming mail for me while I was doing all the hardware banging, though, so no loss in MTA service.) I didn't get the transition done as soon as I'd hoped because...
I had to go up to Sunset Point to "celebrate" my sister's (Sara's) birthday with our grandparents. What joy. That's four hours of my life I'll never get back. Nice and windy up there today. Not toooo bad though (although you'd probably hear differently from a few of the other people in the crowd). Probably about 30mph and gusts around 40 or 50mph.
That' about as eventfull as today gets. Of course, there's still six hours left...
I just wanted to make a note to remind myself about my dealings with "Creative" Labs. I've had correspondance with them all week, with varying answers. For the VideoSpigot, the answer is quite definitive. Since the Spigot was NOT made by Creative, they have no information on it. I'm told it was made by SuperMac Technologies, Inc, a now defunct operation. According to my research, they were acquired by Radius. I've mailed the Radius Vintage product support dept twice now with no answer of any sort (besides from the non-Vintage product support saying that they were forwarding my request to the Vintage dept).
The Blaster is a little more interesting. The first response I got from Creative was that there was no information available on either card. The second (and somewhat unexpected) response says that there's no info on the Spigot, but they do have Blaster info. But, it's only available under an NDA (something I had happened to sign about five years ago for some reason I can't remember). So that info is on it's way. Whether I use it or not is a different story. I can do several things: 1) port the current userspace blaster stuff to the Video4Linux APIs, 2) make a new Video4Linux driver from the NDA'ed info and not release source code and being ever destinted to be excluded from the kernel or 3) the same as 2 except I release the code and hope no one at Creative's Legal gives a damn about a six year old product. Opinions?
Also, some other interesting stuff. I'm not completely in the dark about Spigot. The first response from Creative yielded a zip full of NT 3.5 source code for NT's kernel-level VideoSpigot driver. Looking at that code is scary. I'm not sure I even want to mess with that card. The way it's designed was seriously messing with my mind. And, writing a kernel driver for Linux from that NT code would be both illegal and taxing. It does some pretty wierd stuff with 8k and 128k FIFOs directly from the card into userspace. I'm going to stay a safe distance from that code until I get really bored someday (probably sooner than I would like).
On an unrelated note, my machine (zeta) keeps freezing cold (it's done it twice in the past 24hrs). I was planning on going to 2.1.xxx on this machine anyway, so I may speed up those efforts. It's currently running 2.0.33 (and I'm not going to use 2.0.34 -- I hear it's got some serious aic7xxx oddities). 2.1.104 is out. I'm leaving ihpled at 2.1.103 at least until I get all the configs stable and working. Note to self: when compiling the aic7xxx driver in 2.1, do NOT turn tagged queuing on -- it's broken with Micropolis drives.
I decided that I shouldn't really mix non-ihpled stuff into the ihpled log, so I put this here. Let's have some fun.
As I mentioned, I didn't make it to the DMV to get my driver's permit today. I got up too late and wasn't really conscious until after the place had closed. It will have to wait. I should be doing behind-the-wheel either this coming week or early the next. (Lanée wants me to drive with her, so it will be earlier than most people do it.) This day is turning out to be quite boring. Nothing exciting.
I found a Diamond DVC1000 PCI video capture card for $29, so I'm going to get one of those one of these days. Also, all the token cards should start coming early in the week (ie, Mon). That should be fun.
My dad's really pushing me for delphi's motherboard, but I can't really hurry too much without the Promise card, which is delayed because BofA (pronouced: bo-fa) and VISA are being nasty with their accounting systems causing everything I try to order to be denied. They say they have it all fixed, but, guess what?, I don't believe them. None of these companies will talk to me now because they think I'm a fraud. Damn it! It's amazing (and scary) how someone else's problems can so easily become yours these days. (Who said we have control over our own lives???) But, he can't have that mobo until I get the promise, because I can't freeze ihpled's config enough until I have all the hardware I need! I can't cross anything off the ihpled TODO list yet, but all the lower layers are there.
What I'm looking for right now: a dual (or quad) processor 486 EISA mobo. Yes, that would be an oddity. That's why I want it. That, and I want another SMP board, but I can't afford any more Pentiums.