PREMATURE SPECULATION! [LWN.net]

PREMATURE SPECULATION!

Posted Aug 15, 2009 20:42 UTC (Sat) by michaelkjohnson (subscriber, #41438)
Parent article: A possible change of direction for Foresight Linux

Whoa, Jon!

First, I DO NOT set the Foresight agenda. List participants know that, but readers of LWN might not; your posting this article with this title makes it look like I have influence that I don't. As I posted later in the thread, consider me a gadfly. I'm a minor Foresight contributor, that's all, and for quite some time I've acted as a resident contrarian for Foresight as my main (small) contribution.

Second, I'm speaking ONLY for myself, and not for rPath.

Third, Foresight is already moving away from rPath Linux as a base; it has done so gradually by replacing many elements a few at a time; this is something that Conary makes reasonable and possible. That is fine, including from rPath's perspective -- rPath Linux is only one of the many platforms that rPath makes available in Conary format, and we have ALWAYS wanted that to be the case. From the earliest days of Conary, I've encouraged other distributions to look at Conary; after a while, it turned out that we had to roll up our own sleeves and import the other distributions. I'm not rPath's "Director of rPath Linux", I'm rPath's "Director of Operating Systems". I care deeply about rPath Linux and want to preserve its value; part of that is to not dilute it by trying to make it all things to all people.

Note that moving away from rPath Linux as the base is NOT the same as moving away from "rPath" as a base. If Foresight was no longer based on rPath Linux, it would still absolutely use Conary, rMake, rBuilder, rBuild, and other rPath technology. The question of whether one particular Linux OS platform is the right base OS platform for Foresight is practically orthogonal to using rPath technology.

I proposed this primarily as a thought experiment to help focus discussion on Foresight's core strengths, because I think that Foresight is trying to be too many things to too many people. It has engendered plenty of discussion on the list and in IRC.

"Minor contributor makes provocative suggestion to focus discussion" isn't really front page news.

to post comments

PREMATURE SPECULATION!

Posted Aug 15, 2009 21:44 UTC (Sat) by corbet (editor, #1) [Link] (3 responses)

Whoa Michael.

I posted it as "an interesting set of suggestions." Under the headline "possible change."

Interesting development-oriented discussions are very much front-page news at LWN. My apologies if you didn't want this particular discussion to be quite this visible. I don't think I overstated things, but, then, I seem to be doing well at stepping on toes recently.

PREMATURE SPECULATION!

Posted Aug 16, 2009 0:40 UTC (Sun) by Ed_L. (guest, #24287) [Link]

Well Jon, I don't think you did either. OTOH, if you had kept this off your front page, clueless dweebs e.g. myself would never have heard of Foresight and forever remained blissfully in the dark. Now you have me intrigued: I've been fairly happy with CentOS on my home PC for the last year, but recently find myself pining for the fjords of fedora. I've several alternate F11 partitions, and hope to be able to dual-boot into one of them as soon as I figure out why they both hose my md raid.

(And they ask why CentOS remains so popular...)

Thanks for the site!

PREMATURE SPECULATION!

Posted Aug 16, 2009 1:35 UTC (Sun) by michaelkjohnson (subscriber, #41438) [Link] (1 responses)

Again, I fail in my attempts at humor. It's a good thing that I don't try to make my living in comedy! Sorry.

My (successful) purpose was to provoke discussion within the Foresight development community. (It's been a busy day for me.) My concern with it as a front page LWN article is that as a front page article it loses context, which can give misleading impressions. That's hard to avoid.

Perhaps I should have started out by describing it as a thought experiment, but most of the time when I see that done it ends up leading to bikeshedding. I suppose I could have used the subject line "A Modest Proposal" but I'd expect that to have the same result.

Stepping back, though, away from Foresight -- my modest proposal was certainly possible with Conary; in fact, Conary makes it sane to maintain. The question is not whether it's possible (it is), but whether it is a good match for Foresight's goals. Completely independent of Foresight, I'm still interested in seeing more operating systems packaged with Conary, including but not limited to Fedora. Building them completely from source has some advantages (full application of Conary policy and build rules enhancing consistency, for example); a binary import has other advantages (expediency, consistency with upstream).

Hybrids are even possible -- start with a binary import, then package by package move to building from source where it's useful. Frankly, if any existing Linux distro considers moving to Conary, that's how I'd suggest they start -- working in parallel with existing systems at least to start. For example, Conary not only can import RPM and dpkg, but it can also export tarballs, which can then be used to build RPM and dpkg packages, and that could be enhanced to export more directly to make more use of Conary metadata (for example, automatically creating appropriate install/uninstall scripts instead of having to maintain them explicitly in spec files or debian source packages). I suspect that it's hard to understand why you would want to do that without experiencing Conary, so a binary import is probably the lowest-friction way to get to an apples-apples comparison.

PREMATURE SPECULATION!

Posted Aug 16, 2009 6:09 UTC (Sun) by rahulsundaram (subscriber, #21946) [Link]

Even if you decide not to use Fedora as upstream, I would still be interested in seeing a binary import for comparison.