Menu:

May 28, 1999: Well, I'm here in sunny Phoenix, Arizona. Unfortunately, my computer and the rest of my furniture isn't here yet! For those who are curious, I am now a system administrator and programmer for EST (the BRU guys). More when I have a real computer to type on rather than this crappy laptop!

May 11, 1999: Well, withdraw the job hunting thing, got one! More info when I get there...

May 7, 1999: Well, guys, it's time for me to go job-hunting again! Anybody who needs a good Linux sysadmin/support/etc. type, check out my resume.

April 27, 1999: The cat's out of the bag. VA Research leaked. The papers are at LHS's lawyer's office. Fun fun fun. I fly out to San Jose Thursday afternoon so I can check things out at VA Research on Friday.

April 21, 1999: Fun fun fun. Installing Debian Linux all day.

April 20, 1999: Lots happening that I can't talk about. I'm stressed out to the max. I've now moved my web site back to Tripod while I think about things.

April 9, 1999: Got my 2.0.36 to flash. Sigh.

April 8, 1999: I have the Netier swapping over the network! This little guy ROCKS. I estimate that you can put 6 diskless Netier thin clients per Netier thin server with no problem. Now I have to figure out why I can't get my 2.0.36 kernel to load right off of the flash disk :-(. (2.2.5 worked fine, but I haven't figured out how to make it NFS-swap yet).

April 7, 1999: I can't tell you what I did this morning.

April 6, 1999: Been playing with the Netier thin client/thin server stuff. Works great under Linux. I have a diskless thin client netbooting via DHCP off of the thin server. I had to heavily modify the init scripts though to make this work without a jillion errors. The only irrritance is booting from FreeDOS because the flash disk doesn't have Linux drivers (the flashdisk has FreeDOS on it and uses loadlin to boot the kernel). Now all I have to do is figure out how to swap over the network!

March 1, 1999: OIMS is now usable for everyday inventory management. The only thing that needs addressing is partial shipping. Shipping labels would be a great idea too...

February 9, 1999: I've really been slack, eh? I've put the OIMS stuff on the board. So far only the contact database portion is finished, and that only well enough for use by the inventory portion (i.e., it needs reports and stuff).

November 2, 1998: Added "Cool Systems" to the site.

October 27, 1998: The whole build-a-computer system is now working! See it at http://www.linux-hw.com/config. All that's left is cleanup and prettying up the displays. The data is LIVE and is working. Last thing to do program-wise -- set the transaction cookie in Build.py and read the transaction cookie (if we don't have a transaction number for some reason) in Transaction.py. Also put "method=post" as part of the HTML code generated by the #form tag.

October 3, 1998: Now have persistent transactions going (using MySQL to store the data between CGI calls). Have ported my efforts to Python because the willy-nilly syntax of Perl was driving me nuts. Lack of an equivalent of Python's "pickler" (which can take any arbitrary Python data structure and turn it into something storable in a database, and vice-versa) in Perl was the final straw. Since Python does not have an equivalent of "ePerl" I had to write a DisplayFile.py file which does something similar except you call the CGI script and it calls DisplayFile to display the corresponding HTML page. DisplayFile also takes care of refreshing form data from the contents of the persistent transaction data and etc. I have about halfway tested DisplayFile, the #include and the #variable stuff are working, I now must test the forms stuff. Then comes the CGI<-->Transaction handler (which loads the current transaction state semi-automatically and handles refreshing the persistent data according to what the form just sent us).

September 25, 1998: Have made major modifications to OIMS data dictionary as things move up, including adding a session manager table. The login code is now written. Now working on the session manager/prelude code that goes at the top of each web page. *NOT* using the browser authentication because it sends the user name with each transaction -- meaning I'd have to run the whole thing on the secure server, not just the parts that do authentication (can you say *s*l*o*w?). On a non-secure server there is still the possibility of hijacking a session. Hopefully that will be a remote possibility, because the destination IP address is part of the session state info and the session cookies themselves are pseudo-random and timeout after five minutes. Meaning it'd have to be someone else on your machine with root access, meaning you have bigger problems than session-hijacking. (Well, discounting IP spoofing, which any decent network architecture should take care of with "smart" hubs and lots of switches to segment the network).

September 24, 1998: Put "The Republican Song" on the web site. Added yet more disclaimers as to the offensive nature of the material to certain narrow-minded people with lots of money and no social conscience.

September 22, 1998: Well, I survived Hell Week and even sold a couple of systems. Moved OIMS up on the priority list. Finished the preliminary data dictionary and have written a bit of code (not enough to be usable). The data dictionary is now over in the "software" page.

September 11, 1998: Decided on a name for the new web-based enterprise software: OIMS (Order/Inventory/Manufacturing System). Since it is web-based, it is webOIMS. The initial module (the contact database) should be ready Real Soon Now. It's fun dipping back into database programming using modern tools... I've written over 100,000 lines of database code in the past three years, but all using proprietary and less-than-satisfactory tools.

September 4, 1998: Lots to do while waiting for 300w power supplies to be delivered. 1) Found source for Xeon motherboards and processors, so we will be offering a Xeon-powered system Real Soon Now. 2) Did *NOT* find a source of AMD K6-2 333mhz processors, despite AMD saying they'd dedicated their entire 3rd quarter production to that particular processor. Lots of K6-2 300's, tho. 3) Did *NOT* find anybody who had Super-7 (100mhz Socket-7) motherboards in stock, except for some people who had some off-brand trash motherboards, so even if we had found the processors we would have no motherboards to put them into. (No motherboard worth selling anyhow). 4) There will probably be a two week or so delay in PII-266 and possibly PII-300 deliveries because our suppliers are holding off on orders until after Intel's already-announced price reductions take place September 14. There does not seem to be a shortage of 333, 350, or 400mhz PII processors. We stocked up on PII-266's but large orders could cause us to run out. 5) Don't even ask about Pentium Pros or Pentium MMX processors, Intel has discontinued them. 6) Celeron 300A and Celeron 333 are still not available from our suppliers, and we don't announce vaporware (unlike SOME of our competitors) so we won't announce any systems based on these processors. The Celeron 300A is about $30 cheaper than a PII-266 and overall is about as fast as a PII-275, so it looks like it may be a good processor for our low-end machines (unlike previous Celerons, which we call "celery sticks" around the office because they're about as fast as celery sticks).

September 3, 1998: *FINALLY* getting the hurricane-caused backlog out the door. Magic UPS truck didn't deliver for three days, meaning that we didn't have the parts to put the machines together until this week :-(. I hate hurricanes.

August 29, 1998: Played with GNOME a little bit. It has a ways to go before it threatens KDE. I have the Gnome "panel" at the bottom of my screen (auto-hid) now. GNOME is a lot neater-looking than KDE is, but there's still not enough of it for me to throw KDE off of my desktop. GNOME seems to work best under Afterstep at the moment -- Enlightenment is too simplistic, fvwm2 is okay but I haven't had a chance to configure it to be, well, Gnomish, but right now I'm under kwm 'cause I'm still using KDE.

August 28, 1998: Well, the roof leaks at the office, but none of our stuff was damaged. The ISP in the front office had a large section of ceiling tile fall in but they had covered their stuff with plastic so no problem. We have a bunch of machines lined up ready to go except for missing parts that didn't come in because of the hurricane. Hopefully next week the magic brown UPS truck will come by to deliver the missing parts.

August 27, 1998: Survived Hurricane Bonnie with no problems. Didn't even have to spend much time without power -- the power came back on around midnight after going off at 4pm, and I was grateful because I had a choice between being hot and being wet (the "wet" part being what happens when you open windows during a hurricane).

August 19, 1998: Doing lots of benchmark work on the Behemoths, which is my nickname for the $25,000 mega-servers that we're building for somebody. The "sqlbench" results: 183 seconds on 2.1 with SMP. This is with 10,000 RPM drives on a 6-drive RAID5 array with the GDT RAID controller and dual Pentium-II 400mhz processors, and 512mb of RAM. By comparison, an Alpha 533sx with 128mb of RAM, an Adaptec UW SCSI controller, and a 7200 RPM UW SCSI drive took over 700 seconds to run the same benchmark. Never underestimate the power of fast I/O! I've also detirmined that the 2.1 software RAID0 is faster than the GDT's RAID0, but that the GDT's RAID5 is faster than the 2.1 kernel's software RAID5. My conclusion is that the ECC calculation in the GDT is a) very fast due to the 960's specialized instruction set, and b) probably the result of a better compiler than the GCC compiler. It may even be hand-coded assembly language for all I know. Hard to believe that a 400mhz Pentium-II can't out-calculate a 960 running at a much slower clock rate, but it appears to be true.

August 18, 1998: Got my cats one of those catnip-filled cardboard scratching thingies. I figured they'd sniff disdainfully at it and go back to sharpening their claws on my chair. Instead they went nuts over the catnip and are currently in the process of shredding it. Imagine that -- cats actually using a scratching post!

August 5, 1998: Here I am in Wilmington, NC, 1,000 miles from home. Lots of miles under the belt since the last entry in this log. As I'm sure you have noticed, I'm now working at Linux Hardware Solutions as their tech guy.

See Even Older Old News