Re: [9fans] UN to fund linux for the 3rd world

[dvd@davidashen.net]

>>>It is better than Plan 9 because I don't have to debug its kernel to
>>>do my day-to-day application programming; better than a recent Linux
> 
> what were those kernel bugs?  i can't remember seeing (that many) patches
> on the list or in patches

The patches were not here because it was partly before the patch utility 
had been made working and the list didn't like my ape/diff+ape/patch submittions. Partly after I lost interest in Plan9 as in production system for my programming. Strictly speaking, they are not in the kernel expect for a patch for my specimen of nvidia which didn't like something (I don't remember exactly what, can go
through the patches).

They are in: kfs -- the kfs as it is now and as it was half a year ago cannot
recover cleanly from a filesystem crash. I've modified the code so that it
can. Network libraries -- a few buffer overruns in various buffers which I
stubmled upon due to too long host names in my stress test; don't exactly
remember what else but can try and go through the patches. My iTuner Plan9
box kept hanging  until I found what I thought was a race related to fossil
fs and changed the code so that it does not. It does not hang now.

Smaller fixes include:

 rewriting the code that parses various configuration
files so that I at least understand what is going on -- for example, the timezone
file requires extra blank line in the end if numerical timezone adjustments
are absent; and the zone name is of fixed length (3 characters), while the
time zone I am in (AMST right now) is four characters long. ndb configuration
parses insanely, and experience with Plan9 is actually what made me think
that XML configuration is good -- the syntax is hard too read, but at least
it is predictable -- you don't have to use acid to understand configuration.

My acme mail keeps stopping working after a few days, saying that it
cannot execute a directory, I didn't yet find time to fix it, I will because I
still need Plan9 working as an exotic platform for testing.

Wiki code, both the server and the acme plugin, was unsuitable for any
language but English, and contained many bugs which rendered it unusable -
I rewrote significant parts of the code to make it working.

C compiler's parser gives occasionally error messages which, while indicate
errors, are completely unrelated to errors' sources -- I reported that to the
list, you can browse the source archive.


Overall, my feeling that everytime I start doing something less trivial than ls
(which displays the same file several times on occasion, but it is not a bug,
but simply a wrong implementation of globbing in rc), I have to fix a few
glitches here and there. This is not hard -- I am used to do it in some of
the programs I use, but it is annoying.

The main advantage of FreeBSD is that when I bought that Fujitsu T1120,
I only had to add the identifier of my decoder to ac97.c, tweak ACPI code
a bit so that it turns off power on shutdown (actually I just merged in the
changes from the CURRENT branch); additionally, if my memory does not
betray me, I had a few hard hours forcing mono to install, and finally 
changed the gacutil's code a bit on a nearby Linux machine. That was all,
it happened three months ago, and I didn't touch others' code in FreeBSD
since then; I just use it.

The main problem with Plan9 is that I cannot just use it; it does not work
without fixing.

David