On 2005 Apr 30, at 5:22 PM, Jeff Bachtel wrote: > On Sat, Apr 30, 2005 at 02:30:28PM -0700, Ben Goren wrote: >> As much as I'm sure Theo would love to get rid of gcc and >> friends...damn, that's a big undertaking. I don't think it's the sort >> of thing that would happen at a hackathon. If I had to guess, it'd be >> made the main point of some future release, with little other >> development. You know, the sort of thing that takes up lots of long >> winter nights. It's not likely to be fun, and I get the impression >> that >> hackathons are supposed to be fun. > > *blink* tech@ bitching aside, the GNU gcc project puts a HUGE amount > of effort into improving their compiler. Forking it would have every > chance of leading to stagnation in the OpenBSD project as processors > and optimizations evolve.
There are lots of things I don't know about the inner workings of OpenBSD. Of this, though I *am* sure: OpenBSD will never ``fork'' gcc. Oh, sure, there will probably always be OpenBSD-specific patches against it, or they may lag behind the current version, or what-not. But a fork? I just can't imagine it. Well, *maybe* if they moved gcc to this new version of the GPL everybody is talking about, but I can't really imagine that happening, either. In this distant, dim future I'm envisioning as gcc-less, I would expect Theo to either adopt TenDRA or roll his own compiler from scratch. I have no clue if TenDRA is anywhere close to being up to the task, and writing one's own is quite obviously about as mammoth a project as one could undertake. Notice how I started this paragraph with ``distant, dim future''? All in all, this topic is about as meaningful as us discussing petroleum replacements. Everybody knows that it'll have to happen some day, and perhaps even sooner than anybody really wants--though most would also be overjoyed to wave a magic wand and be done with it. There's a lot of pain between here and there and, while of utmost importance, it's also (currently) about as far from urgent as one can get. (Some day it will be very urgent, unless we discover that magic wand first.) >> This is pure WAG speculation, but I'd guess that the next Open****D >> would be OpenSMTPD [. . . .] > > Replacing Sendmail outright seems iffy at best. Search archives for > when this has been mentioned in the past, and you will get "no way in > hell" replies from Theo. Auditing and partitioning it, maybe. Again, you twist my words, and Theo's too, this time. I did *not* say that I expected a Sendmail replacement any time soon--quite the opposite. Let me put a definite limit on this: I'd bet no more than (a modest) lunch, and only on the condition that I already happened to be in the same city when the bet came due, that, the next time we see an OpenWHATEVER, it's an OpenSMTPD. We may not ever see another OpenWHATEVER, though that bunch just has too much fun hacking for me to imagine that. There may well be something like OpenNTPD that somebody like Henning is quietly working on--and, in fact, this is more likely to be the case than any particular specific thing. Yes, Theo has been quite vociferous about Sendmail replacements. You'll note, however, that in every one of those threads, Theo is responding to people demanding that he replace Sendmail with some specific other MTA. There is no way in hell that Qmail, for example, will ever go in OpenBSD. You'll also note that Theo has never said one word about writing his own MTA. Let me stress that: Theo has been ABSOLUTELY SILENT. He has said NOT ONE WORD. I would be damned surprised if he ever did--unless, of course, it was to say, ``Hey guys, foo@ has done some great work in writing OpenSMTPD for us. Grab a snapshot and help test it for us.'' All the speculation here is MINE. I'm doing a bit of Kremlinology, is all. It doesn't take a genius to notice all of Sendmail's warts. Of the items remaining in /usr/src/gnu, it's got the most license problems. In the past, warty code with undesirable licenses has been a prime target for replacement. All I'm really saying is that Sendmail is the wartiest piece of code left with a problematic license. And why should *you* care whether or not any of this ``scales.'' Managing the project is Theo's worry, not yours, and he's shown himself to be damned capable of doing so. Besides, which do you think is easier: maintaining your own code that you know better than anybody else, or maintaining somebody else's code that you have to work to figure out? And I know from personal experience that, when you get the code right in the first place, you don't have to do anywhere near as much to maintain it. Cheers, b& P.S. As long as we're talking about elephants in the room, it's worth mentioning Perl. Huge code, and all those modules written by everybody and his sister. I don't know which would be worse: writing an OpenPERL or writing C replacements for all that incredible stuff Espie's done to the ports system. Selfishly, I'd rather see OpenPERL...but I really can't see any of this happening until after gcc is gone. Read: not for one helluva long time. b& [demime 1.01d removed an attachment of type application/pgp-signature which had a name of PGP.sig]
