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/2005/4/7/150" onclick="this.href='/lkml/headers'+'/2005/4/7/150';">[headers]</a>聽 <a href="/lkml/bounce/2005/4/7/150">[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><a href="/lkml/2005/4/7/145">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 class="origin"><a href="/lkml/2005/4/7/175">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 08:32:04 -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, David Woodhouse wrote:<br />><br />> On Wed, 2005-04-06 at 08:42 -0700, Linus Torvalds wrote:<br />> > PS. Don't bother telling me about subversion. If you must, start reading<br />> > up on "monotone". That seems to be the most viable alternative, but don't<br />> > pester the developers so much that they don't get any work done. They are<br />> > already aware of my problems ;)<br />> <br />> One feature I'd want to see in a replacement version control system is<br />> the ability to _re-order_ patches, and to cherry-pick patches from my<br />> tree to be sent onwards. The lack of that capability is the main reason<br />> I always hated BitKeeper.<br /><br />I really disliked that in BitKeeper too originally. I argued with Larry<br />about it, but Larry (correctly, I believe) argued that efficient and<br />reliable distribution really requires the concept of "history is<br />immutable". It makes replication much easier when you know that the known<br />subset _never_ shrinks or changes - you only add on top of it.<br /><br />And that implies no cherry-picking.<br /><br />Also, there's actually a second reason why I've decided that cherry-<br />picking is wrong, and it's non-technical. <br /><br />The thing is, cherry-picking very much implies that the people "up" the <br />foodchain end up editing the work of the people "below" them. The whole <br />reason you want cherry-picking is that you want to fix up somebody elses <br />mistakes, ie something you disagree with.<br /><br />That sounds like an obviously good thing, right? Yes it does.<br /><br />The problem is, it actually results in the wrong dynamics and psychology <br />in the system. First off, it makes the implicit assumption that there is <br />an "up" and "down" in the food-chain, and I think that's wrong. It's <br />increasingly a "network" in the kernel. I'm less and less "the top", as <br />much as a "fairly central" person. And that is how it should be. I used to <br />think of kernel development as a hierarchy, but I long since switched to <br />thinking about it as a fairly arbitrary network.<br /><br />The other thing it does is that it implicitly puts the burden of quality <br />control at the upper-level maintainer ("I'll pick the good things out of <br />your tree"), while _not_ being able to cherry-pick means that there is <br />pressure in both directions to keep the tree clean.<br /><br />And that is IMPORTANT. I realize that not cherry-picking means that people<br />who want to merge upstream (or sideways or anything) are now forced to do<br />extra work in trying to keep their tree free of random crap. And that's a<br />HUGELY IMPORTANT THING! It means that the pressure to keep the tree clean<br />flows in all directions, and takes pressure off the "central" point. In<br />onther words it distributes the pain of maintenance.<br /><br />In other words, somebody who can't keep their act together, and creates <br />crappy trees because he has random pieces of crud in it, quite <br />automatically gets actively shunned by others. AND THAT IS GOOD! I've <br />pushed back on some BK users to clean up their trees, to the point where <br />we've had a number of "let's just re-do that" over the years. That's <br />WONDERFUL. People are irritated at first, but I've seen what the end <br />result is, and the end result is a much better maintainer. <br /><br />Some people actually end up doing the cleanup different ways. For example,<br />Jeff Garzik kept many separate trees, and had a special merge thing.<br />Others just kept a messy tree for development, and when they are happy,<br />they throw the messy tree away and re-create a cleaner one. Either is fine<br />- the point is, different people like to work different ways, and that's<br />fine, but makign _everybody_ work at being clean means that there is no<br />train wreck down the line when somebody is forced to try to figure out<br />what to cherry-pick.<br /><br />So I've actually changed from "I want to cherry-pick" to "cherry-picking<br />between maintainers is the wrong workflow". Now, as part of cleaning up,<br />people may end up exporting the "ugly tree" as patches and re-importing it<br />into the clean tree as the fixed clean series of patches, and that's<br />"cherry-picking", but it's not between developers.<br /><br />NOTE! The "no cherry-picking" model absolutely also requires a model of <br />"throw-away development trees". The two go together. BK did both, and an <br />SCM that does one but not the other would be horribly broken.<br /><br />(This is my only real conceptual gripe with "monotone". I like the model,<br />but they make it much harder than it should be to have throw-away trees<br />due to the fact that they seem to be working on the assumption of "one<br />database per developer" rather than "one database per tree". You don't <br />have to follow that model, but it seems to be what the setup is geared <br />for, and together with their "branches" it means that I think a monotone <br />database easily gets very cruddy. The other problem with monotone is <br />just performance right now, but that's hopefully not _too_ fundamental).<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>