Bački Breg this time, but we returned the same day over Horgoš (?). Must be some fuckup on the stamps, someone forgot to switch the date, or switched it in wrong direction, or it's just unreadable and I got it wrong. And also the border authorities don't stamp it in any order, they just find enough blank space and hit it, under any angle.
On 26th in upitig2.prg I changed the way I dimension its window, this is already a 4.8 version. Which was probably in the morning; in the afternoon we had a crash-n-burn at Gemenc, the server crapped.
Serving as a server was the first pentium we ever saw, with, I'd guess, whopping 16M of memory. It was in the office for a couple of weeks, in the newly remodeled summer kitchen, which was a wee hut in the yard, across from Ula's room, where Vilmos and Mihály arranged offices for themselves. It seems Szoftex has hit a sweet spot at the time, there was lots of work with wiring the networks around, including parts of the hospital and lots of other places, so now there was a second car, a ford, again a medium sized diesel with bars on the rear doors, same old tax evasion trick.
And that's where they installed this Acer's altos tower. The big box with first pentium... It had Novell 4.02 on it, where the 2 was supposed to mean that child diseases were over. Some time during september the box went to Gemenc and became their server, and a bit later PolC went live. A few weeks later we got another altos, but that one also didn't hold for long - sold - so we got used not to believe in the existence of a server.
Except the network had its problems here and there. We managed to locate and troubleshoot some of them - the x-ray machines would fuck the signal lengthwise and broadside, and if any write was in progress while radiation was on, that's where babies were made to the signal. They eventually moved the cables, added more shielding to them, and okay now everything will be alright.
Except it wasn't, altos would keep shutting down, reporting "parity error". My session table grew fast and its indexes would get damaged - which I already solved, well wrote a remedy but couldn't fix the cause, I wrote the indexer routine and that part was nearly automated. The hardware guy just kept grumbling "mi o kutya fassa poritsh eror" (which dog's dick poritsch error - pronouncing ty the hungarian way, as a softer version of our ć). Which I didn't get either, it's an error which would appear with faulty memory, but the memories were replaced a couple of times already, no good, that's not the cause. We already stayed there twice until almost midnight, he'd be patching the box and then I'd be patching the tables. And then again the same.
To reduce the damage, I decided to comb away the tables, i.e. at least those two tables where easily multiple writes may be concurrent - with twenty machines on the network, soon to be forty. On 26th I wrote razbacaj.prg ("cast in all directions", i.e. spread), which would go through the existing tables and copy each record into a table with the same structure, just filtering by last two digits of the year of patient's birth. I didn't really read the DoBs, them being contained in the matricular numbers, which are formatted yyymmddnnnnn or thereabouts (pretty similar to our JMBG, aka jebiga broj, the fuckit number), so I'd take 2nd and 3rd digit and split by that. There's only one version of that routine, and it's less than fifty rows long. But it worked the first time, and saved our butts. Temporarily. altos kept falling, but at least the damage was limited and most of the rest could keep working while one was fixed. The hardware guy was pulling his hairs out.
2-VII-2021