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: Michael Frank: Re: Golden rule: don't break userland (was Re: RFC: Starting a stable kernel series off the 2.6 kernel)</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 Michael Frank" href="/groupie.php?aid=10283" /><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/2005/12"> [Dec]</a> 聽 <a class="nb" href="/lkml/2005/12/4"> [4]</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/2005/12/4/71" onclick="this.href='/lkml/headers'+'/2005/12/4/71';">[headers]</a>聽 <a href="/lkml/bounce/2005/12/4/71">[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/2005/12/3/55">First message in thread</a></li><li><a href="/lkml/2005/12/3/63">Adrian Bunk</a><ul><li><a href="/lkml/2005/12/3/65">Arjan van de Ven</a><ul><li><a href="/lkml/2005/12/4/37">Denis Vlasenko</a></li><li><a href="/lkml/2005/12/5/84">Michael Frank</a><ul><li><a href="/lkml/2005/12/5/368">Horst von Brand</a><ul><li><a href="/lkml/2005/12/6/208">Michael Frank</a></li></ul></li></ul></li></ul></li><li><a href="/lkml/2005/12/3/80">Matthias Andree</a><ul><li><a href="/lkml/2005/12/3/81">Otavio Salvador</a></li><li><a href="/lkml/2005/12/3/84">David Ranson</a><ul><li><a href="/lkml/2005/12/3/88">Steven Rostedt</a><ul><li><a href="/lkml/2005/12/3/89">David Ranson</a></li></ul></li></ul></li><li><a href="/lkml/2005/12/3/91">Arjan van de Ven</a><ul><li><a href="/lkml/2005/12/3/94">"M."</a></li><li><a href="/lkml/2005/12/3/167">Matthias Andree</a><ul><li><a href="/lkml/2005/12/3/173">Greg KH</a></li><li><a href="/lkml/2005/12/3/193">"Jeff V. Merkey"</a></li></ul></li></ul></li></ul></li><li><a href="/lkml/2005/12/3/111">Jeff Garzik</a><ul><li><a href="/lkml/2005/12/3/127">Greg KH</a><ul><li><a href="/lkml/2005/12/3/197">Dmitry Torokhov</a><ul><li><a href="/lkml/2005/12/4/134">Greg KH</a></li></ul></li></ul></li><li><a href="/lkml/2005/12/3/157">Matthias Andree</a></li><li class="origin"><a href="">Michael Frank</a></li></ul></li><li><a href="/lkml/2005/12/5/108">Rob Landley</a></li><li><a href="/lkml/2005/12/10/95">Ryan Anderson</a></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">From</td><td class="rp" itemprop="author">Michael Frank <></td></tr><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: Golden rule: don't break userland (was Re: RFC: Starting a stable kernel series off the 2.6 kernel)</td></tr><tr><td class="lp">Date</td><td class="rp" itemprop="datePublished">Sun, 4 Dec 2005 16:37:39 +0100</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody">On Saturday 03 December 2005 19:43, Jeff Garzik wrote:<br />> Adrian Bunk wrote:<br />> > IOW, we should e.g. ensure that today's udev will still<br />> > work flawlessly with kernel 2.6.30 (sic)?<br />> ><br />> > This could work, but it should be officially announced<br />> > that e.g. a userspace running kernel 2.6.15 must work<br />> > flawlessly with _any_ future 2.6 kernel.<br />><br />> Fix the real problem: publicly shame kernel hackers that<br />> change userland ABI/API without LOTS of notice, and<br />> hopefully an old-userland compatibility solution<br />> implemented.<br />><br />> We change kernel APIs all the time. Having made that<br />> policy decision, we have the freedom to rapidly improve<br />> the kernel, and avoid being stuck with poor designs of<br />> the past.<br />><br />> Userland isn't the same. IMO sysfs hackers have<br />> forgotten this. Anytime you change or remove sysfs<br />> attributes these days, you have the potential to break<br />> userland, which breaks one of the grand axioms of Linux. <br />> Everybody knows "the rules" when it comes to removing<br />> system calls, but forgets/ignores them when it comes to<br />> ioctls, sysfs attributes, and the like.<br /><br />WRT sysfs, sysfs is dynamic by design to accommodate <br />individual HW configuration. Thus isn't this really a fault <br />of user-space implementation?<br /><br />><br />> Thus, I've often felt that heavy sysfs (and procfs) use<br />> made it too easy to break userland. Maybe we should<br />> change the sysfs API to include some sort of interface<br />> versioning, or otherwise make it more obvious to the<br />> programmer that they could be breaking userland compat.<br /><br />You might need versions for every entry. I'd go for more <br />documentation on proper use.<br /><br />><br />> Offhand, once implemented and out in the field, I would<br />> say a userland interface should live at least 1-2 years<br />> after the "we are removing this interface" warning is<br />> given.<br />><br />> Yes, 1-2 years. Maybe even that is too small. We still<br />> have old_mmap syscall around :)<br />><br />> Jeff<br />><br /><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: 2005-12-04 17:25 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>