short version:

20:32 <@theo> souterrain: NGPT is abandoned.  now they are flogging NPTL
20:33 <@theo> but none of the Linux systems I can find are using NPTL.  they're all using the old non-POSIX-compliant threads.
21:09 <@dragorn> gentoo, slack 10.2, FC4 all use nptl+tls
21:10 <@theo> these are all gentoo and redhat systems i've inspected
21:11 <@theo> so once again the theory doesn't match the practice
21:29 <@dragorn> If you want to discuss the merits of something or if it's available, I'm happy to talk about it, but if you want to just slag things and ignore facts about it being available in modern versions of things, then have fun on your own
21:33 <@theo> well i think what we started arguing from was, ``once again theory differs from practice''
21:33 <@theo> by which i meant, in theory linux has NPTL on all these widely deployed distributions
21:34 <@theo> while in practice i'm using those distributions, and don't have nptl, because i have the wrong version or the wrong use variable or the right use variable but built from a 2.4 buildhost or WHATEVER
21:34 <@dragorn> Bad argument.  In practice it does.  comparing non-current versions is totally specious.
21:35 <@theo> so i'm comapring theory, linux has the stuff, to practice, *I* don't have the stuff
[...more sophistry...]
21:37 <@dragorn> I don't run it on most of my servers, either.

long version:

20:32 <@theo> souterrain: NGPT is abandoned.  now they are flogging NPTL
20:33 <@theo> but none of the Linux systems I can find are using NPTL.  they're all using the old non-POSIX-compliant threads.
21:09 <@dragorn> gentoo, slack 10.2, FC4 all use nptl+tls
21:09 <@dragorn> I'm sure others do too but those are the ones I've personally encountered.
21:09  * dragorn goes to dinner.
21:10 <@theo> these are all gentoo and redhat systems i've inspected
21:11 <@theo> so once again the theory doesn't match the practice
21:26 <@dragorn> If you matched the version I specified for FC, then yes, it does.
21:26 <@dragorn> Gentoo is unversionable
21:26 <@theo> i didn't match versions.  i looked at systems around me sysadmined by reasonable people.
21:27 <@theo> so, why is <shrug>.  maybe it was added in the last few months.  maybe it is an option (i know in the case of Gentoo there are at least two USE variables and a further conditional on the kernel revision used by the buildhost),
21:28 <@theo> _whatever_.  that's what i'm saying theory vs practice.  the stuff is out there, but all the linux systems available to me are still using non-POSIX threads.
21:28 <@dragorn> FC4 ships with tls integrated, and inseperable.  It's statically linked into core system utils.
21:28 <@dragorn> arguing based on what you have installed vs. known versions is pretty specious.
21:28 <@dragorn> I could produce a 2.2.19 kernel system if you want to play that game
21:28 <@theo> well, then, i guess the reasonable sysadmin i know is not using FC4.
21:29 <@theo> obviously it's very hazy to just look at the systems around you, but i think it also gives some insight.
21:29 <@dragorn> If you want to discuss the merits of something or if it's available, I'm happy to talk about it, but if you want to just slag things and ignore facts about it being available in modern versions of things, then have fun on your own
21:29 <@theo> i spent a year bragging about netbsd's scheduler activations that don't work
21:29 <@theo> no one challenged me
21:30 <@theo> solaris has had posix threads since 1993 and scheduler activations since 1997, and surprise, the solaris system on my desk does have working threads.
21:30 <@dragorn> I'm not even arguing about if it's good or not.  I'm simply telling you known versions of distributions which use it.
21:30 <@theo> and a working debugger and ptrace system for threaded programs.
21:30 <@theo> if i look around at linux boxes run by reasonable people and they don't have NPTL, then that gives you some idea of its maturity
21:31 <@dragorn> Not really.  It gives me an idea of how long ago it went mainstream.  And in any case, it's not particurally relevant to wether or not its shipped
21:31 <@theo> without being a thread expert i can't say anything like ``well with NPTL there is a small chance of deadlock if a thread is suspended while it is holding a futex''
21:32 <@theo> but there are going to be problems like that.
21:32 <@theo> and if, as a solaris sysadmin, i didn't think this way when specifying what software to use on a big production bank web server,
21:32 <@theo> i'd be considered irresponsible.
21:33 <@theo> so yeah i stand by the statement that the thread support in linux although probably better than BSD is immature and isn't getting the kind of professional attention that it really needs to be of high quality.
21:33 <@dragorn> I never said it wasn't immature
21:33 <@dragorn> you seem to be making up arguments that I'm not having
21:33 <@theo> well i think what we started arguing from was, ``once again theory differs from practice''
21:33 <@theo> by which i meant, in theory linux has NPTL on all these widely deployed distributions
21:34 <@dragorn> Bad argument.  In practice it does.  comparing non-current versions is totally specious.
21:34 <@theo> while in practice i'm using those distributions, and don't have nptl, because i have the wrong version or the wrong use variable or the right use variable but built from a 2.4 buildhost or WHATEVER
21:34 <@theo> i don't know what specious means
21:34 <@dragorn> False/incorrect.
21:35 <@theo> but if your versions are so current that none of them are installed around me, i think that is a totally valid point
21:35 <@dragorn> If you decide to not upgrade, and I'm not saying you SHOULD upgrade, it's not valid to argue that "it doesn't have it" just because you didn't elect to use it
21:35 <@theo> ok what i said was that theory and practice differ
21:35 <@dragorn> Wether or not it's stable, or if you should upgrade to it and use it, is a totally different argument
21:35 <@theo> so i'm comapring theory, linux has the stuff, to practice, *I* don't have the stuff
21:36 <@dragorn> Which is wrong.  The practice is, this is shipping code in current distributions.  You may not use them, doesn't change the fact it's shipping.
21:36 <@dragorn> Tht's like calling it a theory that public transportation exists, because I have a car.
21:36 <@theo> it is not wrong.  i don't have NPTL.
21:36 <@theo> you could say it's wrong if my site were run in an unreasonable way, but it isn't.
21:36 <@dragorn> You're putting words in my mouth again.
21:37 <@theo> for example, if you dig into it further and find the version of redhat i'm using has been EOLed by RedHat,
21:37 <@dragorn> I'm not saying a thing about if you SHOULD run nptl
21:37 <@theo> but i don't think it has been.
21:37 <@dragorn> I'm saying that just because you aren't, doesn't mean it's theoretical or that it isn't shipping.
21:37 <@theo> i never said it wasn't shipping.
21:37 <@dragorn> I don't run it on most of my servers, either.
21:37 <@theo> what's the big fucking deal.  in theory ``Linux has'', in practice I use linux and don't have
21:37 <@dragorn> doesn't mean it doesn't exist and isn't in use by anyone who has installed the aforementioned distros, and probably others.
21:38 <@dragorn> the big deal is that IN PRACTICE linux has.  IN PRACTICE you do not use.  Theres nothing "in theory" about it.
21:38 <@theo> IN PRACTICE isn't an abstract thing.  it refers to precisely what I am doing, using Linux in a reasonable way.
21:38 <@dragorn> You know what, never mind.  Have fun.
21:39 <@theo> and it is relevant how agressively they're pushing this out.
21:39 <@theo> they first released it and called it stable in *2002*.
21:39 <@theo> what the fuck have they been doing all this time.
End of Lastlog
--- Log closed Sun Nov 27 21:44:44 2005