CINXE.COM

<?xml version="1.0" encoding="UTF-8" standalone="yes"?> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>LKML: Linus Torvalds: Re: Kernel SCM saga..</title><link href="/css/message.css" rel="stylesheet" type="text/css" /><link href="/css/wrap.css" rel="alternate stylesheet" type="text/css" title="wrap" /><link href="/css/nowrap.css" rel="stylesheet" type="text/css" title="nowrap" /><link href="/favicon.ico" rel="shortcut icon" /><script src="/js/simple-calendar.js" type="text/javascript"></script><script src="/js/styleswitcher.js" type="text/javascript"></script><link rel="alternate" type="application/rss+xml" title="lkml.org : last 100 messages" href="/rss.php" /><link rel="alternate" type="application/rss+xml" title="lkml.org : last messages by Linus Torvalds" href="/groupie.php?aid=1" /><script> var _paq = window._paq = window._paq || []; /* tracker methods like "setCustomDimension" should be called before "trackPageView" */ _paq.push(["setDoNotTrack", true]); _paq.push(["disableCookies"]); _paq.push(['trackPageView']); _paq.push(['enableLinkTracking']); (function() { var u="//m.lkml.org/"; _paq.push(['setTrackerUrl', u+'matomo.php']); _paq.push(['setSiteId', '1']); var d=document, g=d.createElement('script'), s=d.getElementsByTagName('script')[0]; g.async=true; g.src=u+'matomo.js'; s.parentNode.insertBefore(g,s); })(); </script></head><body onload="es.jasper.simpleCalendar.init();" itemscope="itemscope" itemtype="http://schema.org/BlogPosting"><table border="0" cellpadding="0" cellspacing="0"><tr><td width="180" align="center"><a href="/"><img style="border:0;width:135px;height:32px" src="/images/toprowlk.gif" alt="lkml.org" /></a></td><td width="32">聽</td><td class="nb"><div><a class="nb" href="/lkml"> [lkml]</a> 聽 <a class="nb" href="/lkml/2005"> [2005]</a> 聽 <a class="nb" href="/lkml/2005/4"> [Apr]</a> 聽 <a class="nb" href="/lkml/2005/4/7"> [7]</a> 聽 <a class="nb" href="/lkml/last100"> [last100]</a> 聽 <a href="/rss.php"><img src="/images/rss-or.gif" border="0" alt="RSS Feed" /></a></div><div>Views: <a href="#" class="nowrap" onclick="setActiveStyleSheet('wrap');return false;">[wrap]</a><a href="#" class="wrap" onclick="setActiveStyleSheet('nowrap');return false;">[no wrap]</a> 聽 <a class="nb" href="/lkml/mheaders/2005/4/7/145" onclick="this.href='/lkml/headers'+'/2005/4/7/145';">[headers]</a>聽 <a href="/lkml/bounce/2005/4/7/145">[forward]</a>聽 </div></td><td width="32">聽</td></tr><tr><td valign="top"><div class="es-jasper-simpleCalendar" baseurl="/lkml/"></div><div class="threadlist">Messages in this thread</div><ul class="threadlist"><li class="root"><a href="/lkml/2005/4/6/121">First message in thread</a></li><li><a href="/lkml/2005/4/6/121">Linus Torvalds</a><ul><li><a href="/lkml/2005/4/6/129">Greg KH</a><ul><li><a href="/lkml/2005/4/7/166">Rik van Riel</a><ul><li><a href="/lkml/2005/4/7/305">Jesse Barnes</a></li></ul></li></ul></li><li><a href="/lkml/2005/4/6/132">Daniel Phillips</a></li><li><a href="/lkml/2005/4/6/162">Jon Smirl</a><ul><li><a href="/lkml/2005/4/6/168">Matan Peled</a><ul><li><a href="/lkml/2005/4/6/174">Jon Smirl</a><ul><li><a href="/lkml/2005/4/6/189">"Hua Zhong"</a></li><li><a href="/lkml/2005/4/6/241">Christoph Lameter</a></li></ul></li></ul></li></ul></li><li><a href="/lkml/2005/4/6/172">Paul P Komkoff Jr</a><ul><li><a href="/lkml/2005/4/6/243">Martin Pool</a><ul><li><a href="/lkml/2005/4/6/246">Jeff Garzik</a><ul><li><a href="/lkml/2005/4/6/252">Martin Pool</a></li><li><a href="/lkml/2005/4/7/31">Zwane Mwaikambo</a></li></ul></li><li><a href="/lkml/2005/4/6/262">Daniel Phillips</a><ul><li><a href="/lkml/2005/4/7/143">Daniel Phillips</a></li></ul></li></ul></li><li><a href="/lkml/2005/4/7/7">bert hubert</a></li></ul></li><li><a href="/lkml/2005/4/6/221">Jon Masters</a></li><li><a href="/lkml/2005/4/7/11">Paul Mackerras</a><ul><li><a href="/lkml/2005/4/7/30">Arjan van de Ven</a></li><li class="origin"><a href="/lkml/2005/4/7/172">Linus Torvalds</a><ul><li><a href="/lkml/2005/4/7/172">Daniel Phillips</a><ul><li><a href="/lkml/2005/4/7/183">Linus Torvalds</a></li><li><a href="/lkml/2005/4/7/224">Sam Ravnborg</a></li></ul></li><li><a href="/lkml/2005/4/7/272">Dave Airlie</a></li></ul></li></ul></li><li><a href="/lkml/2005/4/7/20">David Woodhouse</a><ul><li><a href="/lkml/2005/4/7/54">Andrew Morton</a><ul><li><a href="/lkml/2005/4/7/59">Paul Mackerras</a><ul><li><a href="/lkml/2005/4/7/66">Andrew Morton</a></li><li><a href="/lkml/2005/4/7/84">Geert Uytterhoeven</a></li></ul></li><li><a href="/lkml/2005/4/7/61">David Woodhouse</a><ul><li><a href="/lkml/2005/4/7/68">Andrew Morton</a></li><li><a href="/lkml/2005/4/7/71">Russell King</a></li></ul></li><li><a href="/lkml/2005/4/7/64">David Vrabel</a></li></ul></li><li><a href="/lkml/2005/4/7/60">Sergei Organov</a><ul><li><a href="/lkml/2005/4/7/81">Matthias Andree</a><ul><li><a href="/lkml/2005/4/7/85">Andrew Walrond</a></li><li><a href="/lkml/2005/4/9/73">David Roundy</a></li></ul></li></ul></li><li><a href="/lkml/2005/4/7/150">Linus Torvalds</a><ul><li><a href="/lkml/2005/4/7/175">Daniel Phillips</a></li><li><a href="/lkml/2005/4/7/176">Al Viro</a><ul><li><a href="/lkml/2005/4/7/185">Linus Torvalds</a></li><li><a href="/lkml/2005/4/7/188">Bartlomiej Zolnierkiewicz</a></li><li><a href="/lkml/2005/4/7/190">Daniel Phillips</a></li></ul></li><li><a href="/lkml/2005/4/8/289">Roman Zippel</a><ul><li><a href="/lkml/2005/4/8/299">Tupshin Harper</a></li><li><a href="/lkml/2005/4/9/79">"Eric D. Mudama"</a></li></ul></li></ul></li></ul></li><li><a href="/lkml/2005/4/7/27">Jan Hudec</a><ul><li><a href="/lkml/2005/4/8/26">Matthias Urlichs</a></li><li><a href="/lkml/2005/4/8/316">Marcin Dalecki</a><ul><li><a href="/lkml/2005/4/9/31">Jan Hudec</a></li><li><a href="/lkml/2005/4/10/206">Miles Bader</a><ul><li><a href="/lkml/2005/4/10/212">Marcin Dalecki</a></li></ul></li></ul></li></ul></li><li><a href="/lkml/2005/4/7/86">Andrew Walrond</a></li><li><a href="/lkml/2005/4/7/304">Ian Wienand</a></li><li><a href="/lkml/2005/4/8/4">Chris Wedgwood</a><ul><li><a href="/lkml/2005/4/8/9">Linus Torvalds</a><ul><li><a href="/lkml/2005/4/8/14">Chris Wedgwood</a><ul><li><a href="/lkml/2005/4/8/18">(H. Peter Anvin)</a></li></ul></li><li><a href="/lkml/2005/4/8/43"> ross@lug ...</a><ul><li><a href="/lkml/2005/4/8/172">Linus Torvalds</a></li></ul></li><li><a href="/lkml/2005/4/8/46">Andrea Arcangeli</a><ul><li><a href="/lkml/2005/4/8/113">Matthias Andree</a></li><li><a href="/lkml/2005/4/8/155">Linus Torvalds</a></li></ul></li><li><a href="/lkml/2005/4/8/48">Marcel Lanz</a><ul><li><a href="/lkml/2005/4/8/82">Geert Uytterhoeven</a></li></ul></li><li><a href="/lkml/2005/4/8/76">Matt Johnston</a></li><li><a href="/lkml/2005/4/12/54">Kedar Sovani</a><ul><li><a href="/lkml/2005/4/12/83">Catalin Marinas</a></li><li><a href="/lkml/2005/4/13/4">Ricky Beam</a></li></ul></li></ul></li><li><a href="/lkml/2005/4/8/107">Catalin Marinas</a></li></ul></li></ul></li></ul></td><td width="32" rowspan="2" class="c" valign="top"><img src="/images/icornerl.gif" width="32" height="32" alt="/" /></td><td class="c" rowspan="2" valign="top" style="padding-top: 1em"><table><tr><td><table><tr><td class="lp">Date</td><td class="rp" itemprop="datePublished">Thu, 7 Apr 2005 08:10:21 -0700 (PDT)</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Linus Torvalds <></td></tr><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: Kernel SCM saga..</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody"><br /><br />On Thu, 7 Apr 2005, Paul Mackerras wrote:<br />> <br />> Are you happy with processing patches + descriptions, one per mail?<br /><br />Yes. That's going to be my interim, I was just hoping that with 2.6.12-rc2 <br />out the door, and us in a "calming down" period, I could afford to not <br />even do that for a while.<br /><br />The real problem with the email thing is that it ends up piling up: what <br />BK did in this respect was that anythign that piled up in a BK repository <br />ended up still being there, and a single "bk pull" got it anyway - so if <br />somebody got ignored because I was busy with something else, it didn't add <br />any overhead. The queue didn't get "congested".<br /><br />And that's a big thing. It comes from the "Linus pulls" model where people <br />just told me that they were ready, instead of the "everybody pushes to <br />Linus" model, where the destination gets congested at times.<br /><br />So I do not want the "send Linus email patches" (whether mboxes or a <br />single patch per email) to be a very long-term strategy. We can handle it <br />for a while (in particular, I'm counting on it working up to the real <br />release of 2.6.12, since we _should_ be in the calm period for the next <br />month anyway), but it doesn't work in the long run.<br /><br />> Do you have it automated to the point where processing emailed patches<br />> involves little more overhead than doing a bk pull?<br /><br />It's more overhead, but not a lot. Especially nice numbered sequences like<br />Andrew sends (where I don't have to manually try to get the dependencies<br />right by trying to figure them out and hope I'm right, but instead just<br />sort by Subject: line) is not a lot of overhead. I can process a hundred<br />emails almost as easily as one, as long as I trust the maintainer (which,<br />when it's used as a BK replacement, I obviously do).<br /><br />However, the SCM's I've looked at make this hard. One of the things (the<br />main thing, in fact) I've been working at is to make that process really<br />_efficient_. If it takes half a minute to apply a patch and remember the<br />changeset boundary etc (and quite frankly, that's _fast_ for most SCM's<br />around for a project the size of Linux), then a series of 250 emails<br />(which is not unheard of at all when I sync with Andrew, for example) <br />takes two hours. If one of the patches in the middle doesn't apply, things<br />are bad bad bad.<br /><br />Now, BK wasn't a speed deamon either (actually, compared to everything<br />else, BK _is_ a speed deamon, often by one or two orders of magnitude),<br />and took about 10-15 seconds per email when I merged with Andrew. HOWEVER,<br />with BK that wasn't as big of an issue, since the BK<->BK merges were so<br />easy, so I never had the slow email merges with any of the other main<br />developers. So a patch-application-based SCM "merger" actually would need<br />to be _faster_ than BK is. Which is really really really hard.<br /><br />So I'm writing some scripts to try to track things a whole lot faster. <br />Initial indications are that I should be able to do it almost as quickly<br />as I can just apply the patch, but quite frankly, I'm at most half done,<br />and if I hit a snag maybe that's not true at all. Anyway, the reason I can<br />do it quickly is that my scripts will _not_ be an SCM, they'll be a very<br />specific "log Linus' state" kind of thing. That will make the linear patch<br />merge a lot more time-efficient, and thus possible.<br /><br />(If a patch apply takes three seconds, even a big series of patches is not<br />a problem: if I get notified within a minute or two that it failed<br />half-way, that's fine, I can then just fix it up manually. That's why <br />latency is critical - if I'd have to do things effectively "offline", <br />I'd by definition not be able to fix it up when problems happen).<br /><br />> If so, then your mailbox (or patch queue) becomes a natural<br />> serialization point for the changes, and the need for a tool that can<br />> handle a complex graph of changes is much reduced.<br /><br />Yes. In the short term. See above why I think the congestion issue will <br />really mean that we want to have parallell merging in the not _too_ <br />distant future.<br /><br />NOTE! I detest the centralized SCM model, but if push comes to shove, and<br />we just _can't_ get a reasonable parallell merge thing going in the short<br />timeframe (ie month or two), I'll use something like SVN on a trusted site<br />with just a few committers, and at least try to distribute the merging out<br />over a few people rather than making _me_ be the throttle.<br /><br />The reason I don't really want to do that is once we start doing it that<br />way, I suspect we'll have a _really_ hard time stopping. I think it's a<br />broken model. So I'd much rather try to have some pain in the short run <br />and get a better model running, but I just wanted to let people know that <br />I'm pragmatic enough that I realize that we may not have much choice.<br /><br />> * Visibility into what you had accepted and committed to your<br />> repository<br />> * Lower latency of patches going into your repository<br />> * Much reduced rate of patches being dropped<br /><br />Yes. <br /><br /> Linus<br />-<br />To unsubscribe from this list: send the line "unsubscribe linux-kernel" in<br />the body of a message to majordomo@vger.kernel.org<br />More majordomo info at <a href="http://vger.kernel.org/majordomo-info.html">http://vger.kernel.org/majordomo-info.html</a><br />Please read the FAQ at <a href="http://www.tux.org/lkml/">http://www.tux.org/lkml/</a><br /><br /></pre></td><td width="32" rowspan="2" class="c" valign="top"><img src="/images/icornerr.gif" width="32" height="32" alt="\" /></td></tr><tr><td align="right" valign="bottom"> 聽 </td></tr><tr><td align="right" valign="bottom">聽</td><td class="c" valign="bottom" style="padding-bottom: 0px"><img src="/images/bcornerl.gif" width="32" height="32" alt="\" /></td><td class="c">聽</td><td class="c" valign="bottom" style="padding-bottom: 0px"><img src="/images/bcornerr.gif" width="32" height="32" alt="/" /></td></tr><tr><td align="right" valign="top" colspan="2"> 聽 </td><td class="lm">Last update: 2009-11-18 23:46 聽聽 [from the cache]<br />漏2003-2020 <a href="http://blog.jasper.es/"><span itemprop="editor">Jasper Spaans</span></a>|hosted at <a href="https://www.digitalocean.com/?refcode=9a8e99d24cf9">Digital Ocean</a> and my Meterkast|<a href="http://blog.jasper.es/categories.html#lkml-ref">Read the blog</a></td><td>聽</td></tr></table><script language="javascript" src="/js/styleswitcher.js" type="text/javascript"></script></body></html>