Current Page -
Last modified Thu May 19 22:27:38 CES 2005
rsc9 before we get into deeper things, could someone tell
me why native awk matters at all?
uriel rsc9: because non-native awk sucks
rsc9: I'm trying to do shell redirection from system(),
and it's just anoying as hell
newsham hey russ, I have a question for you about 9p/auth.
when you read the protocol from the remote I get "p9sk1@domain\0",
but it seems kenfs returns "v.2 p9sk1@..." . What does
that mean? Is this part of the negotiation documented
somewhere?
Major-Wi yeah, like system() uses ksh
uriel rsc9: it just doesn't work as expected
rsc9 so system() uses ksh. any other problems?
Major-Wi i thought the plan 9 shell was rc?
uriel rsc9: return codes for system are also weird
rsc9 "it just doesn't work as expected" isn't very specific.
okay, so system() is different. other things?
Major-Wi exits
uriel rsc9: using ksh to run stuff from awk in Plan 9 is
not what anyone would expect(and again, I haven't managed
to get even that to work properly... but maybe I was
doing something wrong)
Major-Willard: yes, exits is another one
rsc9 okay, system and exits. more?
noselasd I think it boils down to people not liking APE. On
the other hand , it's easy to keep in sync with bwks
AWK :-)
uriel rsc9: I think I had another one, but now I forgot :)
Major-Wi it's not native
newsham if you used rc instead of ksh to run system(..) wouldnt
that require more work when porting non p9 code?
uriel newsham: if I want loonix I know where to find it
newsham uriel: system() is from unix, no?
Major-Wi well, on 9 you'd assume rc
uriel newsham: awk is not just part of ape, awk is a fundamental
tool of the system and should work as part of the system
Major-Wi yeah
rsc9 e already change little pieces of awk rather than use
bwk's distribution as is. utf handling is one area.
i'm happy to hear that we should use our own system()
and exit() too.
uriel newsham: no, in awk you really need to get into the
shell to do many things, specially in "real" awk that
is rather limited in functionality(and I like it that
way)
uriel rsc9: I think I had some nasty problem with something
else, but I'm sorry I forgot what it was, I will let
you know when I remember... :/
rsc9 detailing specific things are just way better than
"i made a native awk; let's use it".
otherwise we end up in the cadt model of system development.
uriel rsc9: ok, I agree
Major-Wi i believe it should ALL be native
uriel rsc9: I was just a bit frustrated as my understanding
is that awk was treated more as a "unix backwards compat"
thing rather than a first class Plan 9 member
Major-Wi is this lunix or 9?
rsc9 awk is a kind of middle ground.
uriel Major-Willard: well, maybe we can get there some time,
let's start with somethin and we will see where we
get
rsc9: yes, I guess it has a bit of a dual role
rsc9 boyd, you can be a purist all you want, but once you
port gs for the 100th time because they released a
new version, you get tired of it.
Major-Wi awk? doesn't that get used to mk the kernel?
uriel rsc9: still, i feel that there is no substitute for
it in Plan 9 world, but then I think Rob never was
a big awk fan, right? :)
rsc9 awk is a plan 9 utility as much as sed is. most of
our awk programs don't use system() so that particular
part isn't as well done. but bwk's awk doesn't do utf
and the plan 9 one does. so there are accomodations.
okay, enough of awk.
musasabi Making more changes than necessary just makes updating
ports more pain - and that time could most probably
be spent on something better.
rsc9 musasabi: exactly.
jmk is not writing a new installer. as far as i know,
no one is.
uriel still, awk is not such a big deal, specially if the
issues with calling external commands from it and exits
are fixed; I was just a bit frustrated the other day
trying to get my awk code to work, so I'm sorry if
I made so much fuss about it
Major-Wi rsc9: ken called me 'an enthusiast'
uriel rsc9: yes, that has been cleaned up, still _someone_
somehow came up with the idea that he was...
(thru some email that no one else has seen)
Major-Willard: I think that quite deffines you :)
rsc9 well don't blame me for that. if you can track it down,
great, but he's not.
uriel rsc9: I don't blame anyone, and I'm sorry if it seemed
like it, I just think that if the process was a bit
more transparent that kind of misunderstandings would
not happen
musasabi What I think would be quite nice would be an easy way
to follow plan9 development - what is changed where
and why.
rsc9 this is really what i wanted to talk about.
uriel yes :)
Major-Wi uriel: i'd pissed off some guy at USG and he asked
ken WTF?
rsc9 as far as following what has happened, /dist/replica/plan9.log
is a great start.
noselasd Well, if someone thought someone was writing a new
installer, all it'd take was a post to 9fans :-)
newsham uriel: uh, you're suggesting that no one do work without
first declaring that they're starting?
uriel I got a list of items I would like to see happening
in the Plan 9 dev community :)
rsc9 boyd: and ken said "he pisses everyone off; don't worry
about it."
(actually i have no idea)
uriel newsham: no
newsham: but it's hard for people that comes new to
know what is going on, or what is the dev model
who is in charge, and so on, patch is great, but it's
not clear who/when/what use it
it's no clear who decides if a patch goes in or not(it's
clear for us, but not for someone coming from outside
that has not follwed things for quite some time)
newsham uriel: I dont find it any better or worse than other
projects (eg. cygwin or freebsd)
rsc9 to the extent that anyone is in charge, it's the people
who work for bell labs. but mainly it feels like more
of a collaboration.
uriel that has some advantages, of course, as sets a barrier
of entry, but I think right now that barrier is a bit
too arbitrary
rsc9 jmk and i are the ones who apply patches.
newsham in most projects an "outsider" would come up with a
patch, send it to the person they think is most appropriate
and either they would add the patch or not.
Major-Wi ACTION thinks an applied patch should (auto) mail
9fans
uriel Major-Willard: not 9fans, but we need a commits list,
I'm going to hack it up soon
rsc9 how about an applied patch appends an entry to a log
on sources.
uriel rsc9: it's easy to patch those things on top of patch,
but not all changes go thru patch
newsham uriel: see /n/sources/patch/applied
__20h__ Boyd, you never joined the NetBSD-cvs ML.
uriel newsham: I know, as I said, I'm going to write an script
that generates emails from that
rsc9 patch/list applied
musasabi rsc9: what do you mean by /dist/replica/plan9.log -
I can only find /dist/replica/client/plan9.log which
is not very helpful.. (sorry for the stupid question)
rsc9 musasabi: that's what i meant.
uriel rsc9: that is not an issue, as I said, it works great,
the only problem is that not all chagnes go thru patch
rsc9 but all changes do go into the plan9.log. if you look
at the changes and can't tell why they were made, then
it's fine to ask on 9fans.
mjl- musasabi: you can track all file changes with it, just
not the description
uriel rsc9: it would be nice if there was a single path,
and the comments are really helpful, not everyone wants
to look thru all diffs...
rsc9: also it's nice to see who did the changes, to
maybe comment/ask that person... but maybe that can
be done with sources now...
rsc9 comments just aren't going to happen. i thought about
it for a while, but it's not the way the plan 9 guys
at bell labs work.
someone familiar with the source could certainly read
the diffs and maintain a changelog. uriel?
uriel rsc9: hehehe..
rsc9: well, then I guess I will try to hack something
that picks up the things from sources...
noselasd I would be *mad* if I had to add comment to every thing
I change in projects at work..
rsc9 the fact of the matter is that if you care about why
things were changed, then you're already reading the
source.
newsham is it really needed though?
uriel rsc9: but it woud be nice to at least know who did
the change, I will look if that can be get from sources
rsc9 ls -lm will show you who pushed it out to sources.
noselasd So I actually understand adding comments may not be
"doable" :.-)
uriel rsc9: not everyone does, I know many people that follows
the commits lists of various projets, and they don't
reall all the src, they just read mostly the comments
to know what is going on
(ask Oksel)
__20h__ Comments? I don't even add such things into the code.
mjl- i'm oksel, but under plan9
rsc9 so fine, then maybe someone maintaining a changelog
would be a good contribution, but it won't be any of
us.
newsham uriel: so if you read through the patch/applied and
miss out on 5 or 10% of the additions, are you missing
that much?
uriel __20h__: code should be slef documenting, changes are
different
mjl- anyway, users could easily make commit logs
rsc9 the bulk of the changes come from patch anyway.
musasabi Usually when following a typical open source project
I just subscribe to the cvs list and get mails about
commits and look at the diff if the message indicates
it is something interesting. Of course one can allways
diff but that is not very nice to look "is this something
interesting".
mjl- once a simple mechanisme to comment on changes has
been mae
newsham if you're actively modifying code, you'll notice if
there's a replica/pull that affects code you're working
on and you'll know the code enough to dif
if you're just trying to get a feel for "whats going
on", its probably not that important if you miss an
item or two
rsc9 okay, i think we've beat this changelog thing into
the ground.
mjl- yups
noselasd musasabi: But what is the usefulness of that ? It's
"nice to know" bot imo not significant..
rsc9 i'd be happy to have someone maintain a changelog on
the side and email us if they want to know more about
a change that wasn't done through patch.
uriel noselasd9: it's more significant for people that are
just starting to get into the system
rsc9 switching to patch acceptance criteria (and feel free
to add this to the wiki).
jmk or i apply patches. acceptance criteria aren't
well-defined, but basically i want to see:
uriel rsc9: yes, documenting that a bit would be very nice
rsc9 1. an explanation of what the problem is
noselasd uriel: Ok - thatks a point.
rsc9 2. a minimal set of fixes, with no gratuitous changes
3. the new code has to look like (style-wise) the old
code
4. if the external behavior changes, document the change
m4dh4tt3 #2 meaning to include no whitespace or formatting changes,
yes?
rsc9 #2 yes.
m4dh4tt3 that makes sense
most other projects i know of have the same requirements
rsc9 if i get patches that don't do those 4, then sometimes
i sorry them
with a note explaining what i'd like to see in a future
patch.
musasabi as a sidenote would patches adding commentation to
existing code be welcome?
m4dh4tt3 gratuitous changes of that nature make it more difficult
for the reviewer to actually review the code
uriel yes, sounds very reasonable, just what we needed
rsc9 but most of the time i just bring the patch up to snuff
myself, noting what i did, and then apply it.
uriel rsc9: do you want to make a wiki page about "how to
contribute"? or should someone else(me?) do it based
on what you said here?
rsc9 o for example i edit almost all the man page changes
that get submitted for english and man page conventions.
i haven't edited the wiki in years. i'd prefer someone
else do it.
uriel rsc9: ok, I will do
rsc9 my first letters of sentences are occasionally getting
chopped off due to an acme irc bug. y'all will have
to cope. ;-)
uriel heh, you should try irc7, seems very popular this days
:)
Major-Wi ACTION agrees with that despite #3, but learnt the
necessity for #3 while hacking the 7th Ed sh
m4dh4tt3 thanks, uriel. i was going to jump in, but i don't
have a drawterm open ATM
Major-Wi yo m4
rsc9 if you create a patch and then use patch/email to give
us your email address, you get email notification when
the patch is applied/sorried.
__20h__ Irc7 only supports one channel.
m4dh4tt3 rsc9: i think we can interpolate ;-)
rsc9 any other patch questions?
Major-Wi deletions?
uriel rsc9: "saved"?
m4dh4tt3 rsc9: thanks for providing those guidelines for the
community. from the patches i've submitted in the past
(before patch), i pretty much gathered all that, but
it's good for the rest of the community to know.
newsham 20h: irc7 is f2f's server/client? if so, you can open
multiple windows in different channels, and you can
still use multiple channels in one window if you dont
bind the window to a channel
rsc9 patch/note saved/whatever-is-there will tell you about
saved.
uriel rsc9: seems a bit weird and redundant, just reject
it and sugest put into sources dir
rsc9 saved is for things that aren't going in just because
they don't fit in/feel like plan 9.
uriel rsc9: yes, but then why wasn't rio-bg saved? ;P
rsc9 sorry is for things that have been rejected. the intent
is to clean up applied and sorry once in a while so
they don't grow without bound.
newsham btw, does patch/create work in acme? the first tiem
I tried it in acme I had issues but I dont know if
it was just the long delay that threw me off or the
bits that make it put the rio window into editing mode
(its been too long for me to recall exactly)
musasabi A simple "how to contribute a patch" example could
be nice in the wiki.
uriel rsc9: my point is, why not just let people maintain
external patches in their own dirs...
rsc9 i actually thought rio-bg was the one i saved. i was
surprised earlier today when i noticed i hadn't.
newsham musa: there's an example in the man page, no?
rsc9 patch/create assumes it can turn on hold mode. it probably
doesn't work well in win windows.
uriel rsc9: if the sources "private" dirs are organized a
bit better, I think that would be a much better place
for things that are not ready, or don't even fit into
the main distribution, but that some people might like
to keep around
newsham that would be good to note in the wiki too. (and maybe
the man page)
uriel yup, I have in my TODO list looking over patch/*; it
could use some polishing
BUGS? :)
Major-Wi deletions
uriel but should be easy to fix, I hope..
rsc9 boyd: fix deletions and i'll apply the patch.
uriel what is "deletions"? sorry, I missed that
__20h__ Patch doesn't support the deletion of files.
uriel ah, I see, yes
musasabi newsham: that does not say anything about adding/removing
files (or directories)
mjl- patching files into existence works
rsc9 deletions are not very common so i'm not very worried.
adding files does work.
i moved some stuff into patch/save
saved
any other patch complaints?
uriel rsc9: I think nothing significant... I would say that
still would be nice if the labs would use it, even
without comments, but well..
musasabi an email notification to a list when something happens
could be nice, but by no means very important.
__20h__ What if I want to patch a saved patch?
uriel (maybe a way to just have it apply automatically pathces
if you have the right perms
musasabi: I will do that
__20h__ There's an typo in the rio-background.
musasabi uriel: thanks.
rsc9 it can't automatically apply patches because of the
way we keep our internal source tree in sync with sources.
uriel musasabi: the only thing is that with the current system
I have to make something that checks both sources/replica
and patch... which is a bit anoying as they look qutie
different
Major-Wi nah auto-application is a bad idea
rsc9 patching saved patches doesn't seem lke a very common
case.
you don't need the check patch. patch is for things
pending on sources.
newsham so the patch/email thing causes automatic notification?
What about mailing a mailing list as well? That might
be useful.
uriel rsc9: one thing, what about keeping public the list
of changes from your private tree to the public one?
(if that is possible)
rsc9 what do you mean?
Major-Wi like i said, mail 9fans
rsc9 mailing 9fans will drive away all the people who don't
care.
newsham major: i dont think everything should go to 9fans
uriel rsc9: jmk mentioned in his email that you guys get
an email every day with all the files that are different
in the distro and in your local tree.
rsc9 e're talking pretty trivial stuff a lot of the time.
newsham but creating a new mailing list (ie. googlegroups.com)
is really easy these days
Major-Wi look we have 250 patches
__20h__ Not googlegroups.com.
uriel rsc9: yes, but it's small things that make the whole
dev model ;)
Major-Wi that"s zip compared to the spam and other crapa
__20h__ Google is the commercial NSA of the future.
uriel __20h__: I'm sure nashi can setup some lists in mordor
rsc9 i don't want to make that list public. if we want the
change to go out, we'll push the file. seeing the file
list doesn't strike me as very interesting.
Major-Wi __20h__: :)
newsham 20h: and patch/applied info is sensitive because?
rsc9 i'd be happy to append the mails to a mailbox file
on sources. if people want to watch it and generate
auto emails from it then fine.
uriel rsc9: ok, I jsut thought it was interesting if someone
was working on something that had pending changes or
something, but well
newsham: good question
rsc9: sounds like a cool idea
rsc9 uriel: that's not the way that list works. if something
has changed in our main tree, it almost always goes
out. people working on little projects keep it in their
own home directories for the most part.
the bulk of the files are config.
__20h__ newsham, it's like who cares about privacy, because
they knew already everything?
rsc9 the only two things sitting in the tree that are major
are software cursor support (waiting to finish vbe)
and usb storage (from /n/sources/rmiller, and he asked
us not to put it in the tree yet)
uriel rsc9: I see, that makes sense, then it would be nice
what interesting work might be lurking in the corners
of home dirs, but I guess that is harder ;)
rsc9 that's things sitting in our internal tree that aren't
on sources.
newsham 20h: i'm not sure what the issue would be with the
nsa (or any other govt organization) seeing any public
mailing list. its a public list.
Major-Wi deletions? how about the 0 mode 0 size file?
uriel rsc9: there is people outside the labs working in vbe(I
guess you know alreay)...
__20h__ newsham, it's about Google -- but that's not the discussion.
uriel rsc9: and I know some people interested in hacking
usb storage
rsc9 i saw some comment about andrey having special access
to insider info in the logs. that's not really true.
we'll talk to anyone who emails us. it's just that
a lot of people seem content to speculate on irc instead
of dashing off an email.
newsham irc rumors are more enticing than reality.
musasabi rsc9: well most newbies don't know who to send the
email ;)
rsc9 they handed the vbe code to me to put into the tree,
and i'm still integrating it. i'm working on putting
it in aux/vga instead of 9load. if someone wants to
help, feel free to email me.
uriel rsc9: ok, sorry, that is my fault, but well, one doesn't
want to bother you guys too much with obvious questions...
rsc9 musasabi: then it's up to you old-time irc hands to
point them in the right direction. or you email us
instead.
uriel rsc9: if a contact point was documented somewhere it
would be nice, can we put your email in the wiki as
"contact point"(maybe better 9fans)... which brings
us to plan9-dev list...
rsc9 uriel: people who want to hack usb storage can look
in /n/sources/miller and then contact richard with
changes. he's heading that up.
uriel I understand that some people don't feel confortable
discusing serious stuff in 9fans due to all the noise,
but it seems that plan9-dev failed...
rsc9 boyd: treating deletions as size 0 files (no need for
mode 0) sounds fine to me.
Major-Wi well, i like overkill
uriel rsc9: do you think a low-noise forum is needed for
dev discusion?
rsc9 i'm glad plan9-dev failed. i don't like the dichotomy.
i refused to subscribe for quite a while, although
i wish i'd been there to help with designing 9p2000.u.
i think 9fans is a fine place for dev discussion. it
happens occasionally.
Major-Wi and it saw your can't read it and there's nothing in
it
it says
uriel rsc9: ok, I see, I guess people will have to put up
with 9fans then, the problem I see is that lots of
disucion seems to happen in private email instead..
:/
which leaves most people out of the loop
rsc9 mode 0 files are a pain because you can't even open
them.
Major-Wi and /n/sources? how does one get a dir?
uriel Major-Willard: I was going to get there next
rsc9 that will happen on a dev-only list too. the only reason
to have a dev-only list is that people on 9fans don't
want to see the dev chatter, which i very much doubt.
uriel ;)
musasabi btw would making a common readline function (basically
the readln function which is copy/pasted in many places
make sense? (just found myself copying the code and
I could submit a patch if such a thing would be wanted
and I knew the correct lib for it)
uriel rsc9: I thought it was the other way around, people
interested in dev wasn't interesed in 9fans noise,
so they used priv email instead,,, I guess I was wrong
again :)
noselasd 9fans doesn't seem to be the most overloaded list.
Dev discussion there would fit fine, no ?
Major-Wi READLINE? are you MAD?
rsc9 plan9-dev got created because of the ridiculous amounts
in spam back in july. that's solved.
uriel musasabi: over my dead body
+quintile
__20h__ Boyd, there's no GNU inside. ;)
musasabi Major-Willard: not *nix readline.
rsc9 you don't need readline unless you want to read a password.
uriel rsc9: good, then that is solved
rsc9 if you're reading from a console, read() returns one
line at a time.
noselasd musasabi: mouse around :-)
Major-Wi i don't need readline() i have hold mode
uriel rsc9: souces dirs..
rsc9: I documented in the wiki that to get a sources
dir one should email you or rsc(I hope you don't mind
:))
rsc9 the usual thing with sources dirs is to mail me or
jmk and we'll create it. boyd is an exception.
uriel rsc9: the thing is that there is no criteria to who
gets one dir and who dosn't
rsc9: that dosn't seem fair to me
musasabi rsc9: but I don't want to assume stdin is a console
- and the existing code seems read one char at a time
till it gets to the end of the line..
uriel (actually it seems plain stupid to me0
rsc9 jmk wasn't feeling very kind when boyd asked.
uriel rsc9: well, are you feeling more kind today? :)
Major-Wi rsc9: should be a fortune "<rsc9> the usual thing with
sources dirs is to mail me or jmk and we'll create
it. boyd is an exception."
rsc9 ;-)
i'll make boyd a directory but first i want to put
all the directories in a subdirectory
instead of the root.
uriel I personally would be upset if someone that has done
so much for Plan 9 was left out, I think it's just
_wrong_; boyd might be a difficult person, but that
is no excuse not to let him contribute
rsc9 choices for names? /n/sources/users/foo? /n/sources/userstuff/foo?
uriel rsc9: contrib?
rsc9 i'm not getting into a discussion about boyd.
Major-Wi usr
rsc9 there's already a contrib.
usr is home directories which these aren't.
__20h__ home :P
Major-Wi they sort of are
rsc9 contrib is good.
uriel rsc9: yes, that needs to be cleaned up a bit, there
is "contrib", "extras" "cvs", ... and so on, quite
messy..
mjl- there already is a contrib... yeah
quintile community ?
Major-Wi no
uriel quintile: too long :)
Major-Wi contrib
musasabi users and contrib sound nice.
Major-Wi or wip
uriel vt3 and I volunter to keep an index of the stuff in
contrib then
rsc9 great.
uriel rsc9: some guidelines for how to use sources dirs,
similar to the ones you gave for patch would be very
apreciated
(mostly to keep some order)
__20h__ What's the 9grid dir for?
uriel (and I guess somethign like http://plan9.bell-labs.com/9grid/AUP.html
can be linked from the wiki)
__20h__: good question
noselasd Ick. did irc7 swallow a line of mine :-
uriel (and links into my next question, that is what is up
with *.grid.bell-labs.com.)
rsc9 i don't know anything about grid. ask andrey.
__20h__ Ok.
rsc9 okay, i created stuff in /n/sources/contrib. please
move your fiels.
uriel rsc9: what about *.grid.bell-labs.com then? quintile
said it's still up and running, but there is no way
to get new accounts?
rsc9 i don't know anything about grid. ask andrey.
i believe new accounts only go to people contributing
resources. but i know nothing. it's presotto and ron's
baby, and it's mostly stalled.
uriel ok, will do then, I thought he didn't know as he was
around when it was discused
well, I was asking because all the other 9grid.??-s
poping up over the wrold..
rsc9 i don't know anything.
uriel OK, I will nag f2f when he is back :)
rsc9: about p9p and merging the docs..
rsc9 about sources/contrib, it's for posting software for
plan 9 users. if you use up too much disk or post crap
we'll probably do something.
uriel Oksel: and I ahve looked at it, and it looks nasty..
the changes are too many and too hard to check which
ones should go and with ones are p9p specific..
m4dh4tt3 rsc9: the perms on my dir in contrib are incorrect
:-)
rsc9 fixed
m4dh4tt3 thx
Major-Wi methinks one dir is missing
uriel my proposal is to merge both documentation, and have
an extra section for p9p specific pages, and an extra
section inside pages that are different documenting
the diferences for p9p
rsc9 no.
uriel rsc9: ok, but I tell you, merging back and forth is
unpractical..
rsc9 i don't want p9p to pollute the main distribution.
changes aren't frequent enough to merit that.
uriel I think that you will find it harder and harder to
keep them in sync...
rsc9 if we can get them in sync to start, then keeping them
in sync is easy.
uriel well, if people wants to start working on the Plan
9 docs, how will you merge back? it goes both ways..
rsc9: it's hard for people that might have access to
one but not the other
rsc9 i know when the last time i merged was, and occasionally
i run diff to see what's happened on sources in the
last (say) 3 months and then do it that way.
everyone has access to the dump on sources and to the
cvs history for p9p. both let you diff by date.
the trick is getting them up to date.
that is, in sync the first time.
regardless of what solution we decide on going forward,
that initial work needs to happen.
uriel rsc9: ok, we will try, but seems like a tedious and
futile task from our POV..
rsc9 i don't see why it's hard to figure out, given some
diffs, whether they apply to plan 9.
the way i see it is this.
the goal is to bring the plan 9 man pages up to date.
i already identified some places
where they were out of date and fixed it in p9p when
i did the initial p9p man pages.
diffing p9p and plan 9 man pages just points out some
places where you need to update the plan 9 pages.
i'm not saying you should diff the pages that have
changed dramatically, like intro(1)
uriel yes, but some times it's hard to tell..
mjl- i've only looked at acid, and it's a bit of a problem
that i've never used it beyond lstk()
so a bit hard to know what applies to both now
uriel you have to actually look thru the diff to see which
changes are p9p only and which are p9...
mjl- since i also don't use the p9p version
rsc9 acid is a tough one. i'd skip acid.
mjl- guess that's a problem of too little experience...
ok
will look further then. already have some scripts that
show diffs btw (if anyone wants them)
rsc9 if you diff all the pages and make a big file with
all the diffs, you should be able to pick out the easy
ones.
once you do the easy ones, feel free to email me to
ask about the hard ones.
uriel rsc9: another advantage of keeping a single copy of
the docs is that the implementation differences are
clearly documented somewhere, that way they can be
keept track of, and hopefully be reduced to a minimun
over time
rsc9 i agree with you in principle. eventually i'll probably
use cvs branches so that the p9p cvs is essentially
keeping track of the diffs between the two.
of course, it's naive to believe that all the implementation
differences expose themselves in documentation.
uriel ACTION would recomend using something other than cvs
to keep track of branches, but that is a detail ;)
rsc9: yes, but it's better to try to keep track of
them than just ignore them
newsham whats wrong with cvs branches?
__20h__ dho did some sort of new "cvs" -- no code yet.
rsc9 let's not talk about version control. more cadt crap.
uriel newsham: merging patches back and forth from on branch
to another is quite nasty
rsc9: yes :)
rsc9 i'm not a power user so cvs is fine. and i know how
to use it, crappy though it is.
mjl- rsc9: that's what i was trying with acid too (looking
into the code)
newsham <- likes CVS and their branch management (overall,
some complaints of course)
rsc9 don't try to diff the acid code.
mjl- will do that when writing docs (which is why i come
up with these usage patches ;))
newsham 20h: have you guys looked at larch/arch ?
uriel newsham: you are crazy ;P
mjl- arch... bazaar... yay
noselasd Oh - SCM fest. (/me brings the popcorn) ( *SIGH*)
uriel newsham: I have, lets leave it for antoher time
rsc9 brb. have fun talking about version control.
uriel rsc9: no, no, we gave up already!
mjl- will go for a snack :)
uriel ^_^
__20h__ newsham, It's not my idea. Replica and Patch is enough
for me.
noselasd Please. cvs vs whater. Insignificant.
newsham arch has some nice properties (and seems to be tool
oriented)
noselasd whatever*
musasabi arch is quite complex etc
musasabi point.
rsc9: So refactoring copy-paste code into common code
is generally not a desirable change?
__20h__ And who implements metadata in 9p? :P
uriel newsham: we can argue over arch anothter time, lets
not scare rsc9 off ;)
newsham i have nothing to argue about..
rsc9 back
uriel newsham: sorry, my english sucks, I mean "discus"
rsc9 musasabi: don't understand the question
oh, if you're talking about readline, you should be
able to call read.
how often do you read passwords anyway?
musasabi rsc9: readline (readln and probably other names), e{malloc,free,realloc}
and probably others.
__20h__ auth_proxy();
uriel you could use a gui app for reading pass...
__20h__ Factotum does the password management.
uriel rsc9: it would be interesting to know more of what
are your plans for p9p
__20h__: yup
mjl- emalloc and stuff has already been discussed, read
the archives
rsc9 i don't really have plans. as i need more software,
i convert it. but i'm at a stable point now.
uriel rsc9: I guess I was thinking more from the POV of bringing
both code-bases closer..
__20h__ dtLinux - stable
uriel (I have heard rumours that the venti code in p9p is
more uptodate than in p9...)
rsc9 one thing on the horizon is adding a c front end compiler
so that the sources look more the same.
the sources already look almost identical (acme is
the big exception but even those changes are simple
if tedious)
newsham c->c ?
rsc9 yes
musasabi mjl-: actually I was the one proposing it in the archives
;)
rsc9 http://swtch.com/usr/local/plan9/src/cmd/venti
uriel yes, that is a problem, hopefully I will get charless
to fix up kencc some time this century and the OpenBSD
guys will take care of the porting
rsc9 no no no. i am not putting a real compiler in. just
c->c.
uriel rsc9: just curious, why not?
-pperez
musasabi 8c can be compiled on *nix, but the output is just
not very usable.
--------- 5 mins
rsc9 because porting a compiler requires making it work
with all the system libraries and the like. it's too
much work.
the p9p goal is to play well with unix.
uriel rsc9: I see your point, still I hope some day someone
does it
quintile rsc9: the new venti include auth?
newsham musasabi: wouldn't be that hard to fix it to emit useable
(ie. elf) binaries (some arches already have a linker
flag)
rsc9 besides, who really wants to keep up with all the library
conventions on sunos.
emitting elf is easy. emitting useful debugging is
much harder.
i have an 8c that can create hello, world.
Major-Wi ELF -- yuk
newsham blah dwarf.
uriel rsc9: I'd be interested in seeing that(even if trivial)
rsc9 i have yet to hear a good idea about how to do auth
in venti.
uriel rsc9: what about the venti rewrite
(which I found out about by reading the notes file
in rejected patches....)
rsc9 i'm in the middle of rewriting venti. it's a lot faster,
perhaps a little buggier, and needs a bit of cleanup
before it will be ready for prime time.
any other projects people care about?
newsham wouldnt even a very limited amount of auth be significantly
better than no auth?
uriel rsc9: just curious, but don't you think that this kinds
of cases the Open Source mantra of "release early,
release often", could help?
noselasd What projects are there - (at the labs , if any) ?
rsc9 newsham: let's postpone the venti auth. i can't do
multiple conversations.
newsham ok
Major-Wi i have a cool idea for venti, but it's an expensive
venture
quintile projects - cross domain auth between servers (for a
grid).
Major-Wi too complex
rsc9 uriel: release early, release often only works if you
have time to keep up with the submitted changes. i'm
focusing on something other than venti.
uriel rsc9: ok, I thought it could help with testing, and
maybe someone could pick up if you dont' have the time
to maintain/finish it, but well..
rsc9 the venti code is available if you want to hack on
it. it's in the same place as plan 9 ports. module
name is venti instead of plan9.
uriel quintile: I think that was discused at OSDI and rejected,
the current model seems good enough, otherwise I would
recomend you to talk with charles about the Inferno
auth model..
newsham uriel: so far nobody's really picked up ron's xen stuff
if you're looking for stuff to pick up...
uriel rsc9: ok, thanks
newsham: I know, I'm not saying that it will hapen,
I'm just saying that it costs nothing to give people
a chance
quintile uriel: yep, just interested in russ's opinion.
rsc9 lucho helped me find some things. he emailed me.
uriel newsham: didn't kuroneko pick up your spark code?
__20h__ Are there plans to have more than the PC-distribution?
rsc9 as for venti auth, it's just not a priority.
uriel newsham: and didn't ericvh pick up v9fs even if after
some years?
newsham uriel: so far three people have approached me interested
in sparc v8 stuff. I dont really know what has become
of their work.. they may still be working on it.
rsc9 calling it a pc distribution is a little misleading.
it's a full distribution, it just only installs on
pcs.
uriel newsham: the thing is that maybe you get something
useful out of it, maybe not, but the chance is worth
it, I think,
anyway, just to be anoying, what about 9load... ;P
rsc9 what about it?
uriel is someone working on it?
rsc9 as far as i know, not much is happening with 9load.
uriel I also have heard(from third hand) of a rewrite of
9load..
rsc9 vbe support got contributed for 9load, though i plan
to move it out.
jmk is working on an amd64 port, which includes cleaning
up 9load some, but it's not a from-scratch rewrite.
uriel ok, I guess that leave us free to fix the anoying IDE
handling... :)
rsc9 what annoying ide handling?
uriel rsc9: any chances we will get any of that cleaning
up?
rsc9 any of what?
uriel rsc9: if you don't have your ide devices in the right
magic convination it wont work
rsc9 huh?
uriel rsc9: 9load clean up...
rsc9: ever tries to install having your CD-ROM as secondary
master?
m4dh4tt3 yeah
rsc9 the 9load cleanup will come out when the amd64 port
is ready. i believe the changes to 9load are pretty
minimal so don't worry about it.
uriel rsc9: its' a very commonly reported problem
m4dh4tt3 i do it all the time
rsc9 does the kernel work in that situation?
m4dh4tt3 sure does
noselasd uriel: works fine. You just need to type the magic
sdD1!... and so on :-)
uriel rsc9: it should, I would say... but IIRC it has also
some problems
noselasd9: nope
rsc9 ide probing is black magic. the sdata there is supposed
to be similar to the sdata in the kernel (dma is yanked
out), so feel free to compare and fix.
m4dh4tt3 all you have to do is specify a different drive to
load the kernel and the root fs
noselasd uriel: funny, cause I did that yesterday though..
uriel noselasd9: it doesn't work, and I have gone thru that
problem at least three different times, the people
at the 9con can tell you, we were a room packed with
plan9 users and no one could get it to work until we
changed the ide config
newsham m4d: isnt that because of the plan9.ini, not the 9load?
rsc9 ide probing is black magic. some systems have buggy
bioses too.
m4dh4tt3 newsham: iirc, yes
uriel rsc9: yes, that too :(
rsc9 if you want to go after this bug and fix it, please
do. trust me -- it won't go away on its own.
m4dh4tt3 ide probing is indeeed black magic. i've seen some
really bizarre stuff
uriel rsc9: ok, thanks
drawterm..
who maintains it? f2f says he doesn't, even if he does...
rsc9 i hate drawterm. andrey is doing a good job maintaining
the 9p1 drawterm.
uriel I hate dt too, but it's a fact of life..
rsc9 brucee is busy with other stuff though i think he might
have a good 9p2000 drawterm at some point.
dt2k is worth using, and if it didn't crash once in
a while i'd install it.
eventually i want to replace drawterm with a collection
of programs in plan9port.
uriel would be nice if f2f's changes were in the main distro,
and handled with patch, but f2f won't do it because
he says "I'm not the maintainer", so he basically maintains
a fork
rsc9 but progress is slow. i do have ssl working though.
uriel rsc9: where is dt2k?
ACTION has never seen it, but I know f2f has got the
code for some ancient version of it somewhere..
f2f if you notice, uriel, the binaries have disappeared
from the main distro with a small note pointing to
ucalgary.
uriel f2f: ah, that is new, and I don't like it, but well...
f2f i think rsc just doesn't want that mess in p9
uriel would be nice to have it in sources at least
contrib/f2f?
f2f i'm fine with ucalgary until a better replacement comes.
this is where the dt2k code is too, at least the latest
version i have
rsc9 andrey is in the drawterm group now.
uriel cool
f2f rsc, do you want me to merge the dt changes?
rsc9 the 9p1 changes? yes.
f2f ok, will do
rsc9 brucee is swamped with other work. he won't mind.
uriel ACTION would like to know more about dt2k and what
is going on with it :)
f2f but i'll keep the binaries on the web
rsc9 dt2k works well enough that i use it every day to apply
patches.
it crashes occasionally (some x problem) and i restart
it.
i never got around to a windows port.
i don't believe it's the right approach so i am loathe
to spend time debugging it.
yes please keep binaries on the web.
uriel rsc9: I think I know many people that would apreciate
dt2k even without windoze port
(and I know others that might fix the windoze port)
rsc9 google the 9fans archives for dt2k. i put it out there.
uriel ok, that is the latest version then?
rsc9 sure.
Major-Wi drawterm is pretty solid
uriel ok, will do..
f2f http://pages.cpsc.ucalgary.ca/~mirtchov/p9/dt2k.tgz
uriel f2f: that is the latest version too?
f2f i got this from skip and made it compile cleanly under
X
uriel I see...
f2f that's the latest i've heard about
rsc9 do you use it? does it crash?
f2f it crashes, yes. i don't use it because the alpha blending
is messed up (BGR vs RGB i believe)
and it's a bit slower
axelB (what about "replace drawterm with a collection of
programs in plan9port" and windows?)
Major-Wi BGR?
rsc9 what about it?
uriel axelB: ah, fuck windows! (sorry, could not restis ;P)
axelB just wondering. p9p doesn't do windows, I thought
noselasd uriel: can you ever ? :)
uriel noselasd9: if you knew...
rsc9 eventually plan9port will work on windows. don't care
enough right now. very happy with my mac.
axelB ok. fine
__20h__ Cygwin?
uriel axelB: buy a mac mini, they are cheap ;)
Major-Wi err, p9p port to windows is a major, hideous task
rsc9 it's not any more work than the original unix work.
i've done it before.
m4dh4tt3 i miss the mac i had for my previous job
uriel I guess another issue is p9p integration with v9fs
and similar, but without eric or lucho around I think
better leave that for other time
rsc9 my position is that p9p will not become dependent on
v9fs -- it needs to work on systems without v9fs --
but i'm perfectly happy if lucho and eric need hooks.
uriel rsc9: sounds reasonable..
rsc9 we put one hook in already. auth_proxy tries to open
/mnt/factotum/* before switching to the p9p factotum
socket.
musasabi Is there a reason plan9 headers don't have #ifdef protection
agains multiple #includes ? (stdio.h mainly - some
headers have function prototypes with FILE* and making
sure stdio.h is included just once adds many trivial
(unnecessary) changes.
uriel musasabi: that is a well known one, use sane headers
rsc9 don't include headers twice.
uriel exactly
rsc9 if you're compiling legacy code, then use ape. the
ape stdio.h is protected.
uriel musasabi: feel free to add to the faq
rsc9: BTW, could you update the running version of
the wiki, I really need it to make the FAQ page useable..
Major-Wi Is there a reason plan9 headers don't have #ifdef protection
agains multiple #includes ? -- YES, YOU are 'sposed
to get that right
musasabi ACTION uses the "not my code" and "uriel will flame
me to death for APE" arguments.
rsc9 google for "follow the simple rule" pike
if it's not your code, you should use ape.
uriel or fix it :)
__20h__ Or delete it.
Major-Wi rm is a swell prgramm debugger
musasabi nontrivial socket code + APE was not very nice so I
am trying to live with 8c -p.
uriel rsc9: I got some changes to the templates of the wiki,
I haven't sent them because I have seen that the ones
in the distribution and the ones in the plan 9 wiki
are not in sync anyway...
m4dh4tt3 when fixing it means re-writing it, the probability
that it will get done approaches zero
rsc9 uriel: i think i restarted wikifs
uriel rsc9: let me check again...
[...]
rsc9 gotta go. forgot about another appt.
