{"thread":{"id":"15138","subject":"Re: [kernel.org users] README and ChangeLog files","startedAt":"2008-08-21T15:35:52Z","lastAt":"2008-08-22T23:09:52Z","messageCount":5,"participants":["Linus Torvalds","Petr Baudis","Michael J Gruber","Junio C Hamano","Jeff Garzik"],"isPatch":false,"patchVersion":null,"patchTotal":null},"messages":[{"id":"87999","messageId":"alpine.LFD.1.10.0808210830080.3487@nehalem.linux-foundation.org","threadId":"15138","inReplyTo":"48AD03BD.9000909@zytor.com","subject":"Re: [kernel.org users] README and ChangeLog files","fromName":"Linus Torvalds","fromEmail":"torvalds@linux-foundation.org","sentAt":"2008-08-21T15:35:52Z","receivedAt":"2008-08-21T15:35:52Z","isPatch":false,"sender":{"key":"torvalds@linux-foundation.org","avatar":"https://avatars.githubusercontent.com/u/1024025?v=4"},"body":"\n\nOn Wed, 20 Aug 2008, H. Peter Anvin wrote:\n> \n> Personally, I think this grand renaming was a huge step backwards; an attempt\n> to be bug-compatible with other SCMs.  The only people I personally ever saw\n> complaining about the situation in the first place were people who were\n> fanbois of some other SCM.\n\nHeh. It's true that the push came from people who were used to the \"scm \nxyz\" model, but it's also true that git basically made the situation in \n/usb/bin even worse. \n\nYou _can_ actually have the \"best of both worlds\" (if you really do want \nto continue the \"git-xyzzy\" usage) by doing\n\n\tPATH=$PATH:$(git --exec-path)\n\nalthough there has been noise about removing the built-in commands \nentirely even from there. I'm cc'ing the git mailing list just to bring \nthe point up - I'm the one who actually championed removing the \n\"unnecessary\" hardlinks, and it seems that nto doing so was the right \nchoice.\n\nOne of the reasons that the dashed format is being removed is a real \ntechnical one, though: git aliases. They never supported the dashed \nformat, since they never were real executables (that's the whole point of \nan alias, after all).\n\nSo even with the above path thing, you'll still have to use the spacey \nversion for any aliases you use.\n\n\t\t\tLinus\n"},{"id":"88001","messageId":"20080821160113.GM10544@machine.or.cz","threadId":"15138","inReplyTo":"alpine.LFD.1.10.0808210830080.3487@nehalem.linux-foundation.org","subject":"Re: [kernel.org users] README and ChangeLog files","fromName":"Petr Baudis","fromEmail":"pasky@suse.cz","sentAt":"2008-08-21T16:01:13Z","receivedAt":"2008-08-21T16:01:13Z","isPatch":false,"sender":{"key":"pasky@ucw.cz","avatar":"https://avatars.githubusercontent.com/u/18439?v=4"},"body":"On Thu, Aug 21, 2008 at 08:35:52AM -0700, Linus Torvalds wrote:\n> One of the reasons that the dashed format is being removed is a real \n> technical one, though: git aliases. They never supported the dashed \n> format, since they never were real executables (that's the whole point of \n> an alias, after all).\n\nOn the other hand, with the dashed form you can easily use shell\naliases. Sure, there are differences, like that they are set per-session\nat most, not per-repository; but if you actually define different\naliases with same names in your repositories, that seems like shooting\nyourself in the foot elaborately most of the time.\n\n\t\t\t\tPetr \"Pasky\" Baudis\n"},{"id":"88004","messageId":"g8k5b6$32a$1@ger.gmane.org","threadId":"15138","inReplyTo":"alpine.LFD.1.10.0808210830080.3487@nehalem.linux-foundation.org","subject":"Re: [kernel.org users] README and ChangeLog files","fromName":"Michael J Gruber","fromEmail":"michaeljgruber+gmane@fastmail.fm","sentAt":"2008-08-21T16:32:36Z","receivedAt":"2008-08-21T16:32:36Z","isPatch":false,"sender":{"key":"git@grubix.eu","avatar":"https://avatars.githubusercontent.com/u/233215?v=4"},"body":"Linus Torvalds venit, vidit, dixit 21.08.2008 17:35:\n> \n> On Wed, 20 Aug 2008, H. Peter Anvin wrote:\n>> Personally, I think this grand renaming was a huge step backwards; an attempt\n>> to be bug-compatible with other SCMs.  The only people I personally ever saw\n>> complaining about the situation in the first place were people who were\n>> fanbois of some other SCM.\n> \n> Heh. It's true that the push came from people who were used to the \"scm \n> xyz\" model, but it's also true that git basically made the situation in \n> /usb/bin even worse. \n> \n> You _can_ actually have the \"best of both worlds\" (if you really do want \n> to continue the \"git-xyzzy\" usage) by doing\n> \n> \tPATH=$PATH:$(git --exec-path)\n> \n> although there has been noise about removing the built-in commands \n> entirely even from there. I'm cc'ing the git mailing list just to bring \n> the point up - I'm the one who actually championed removing the \n> \"unnecessary\" hardlinks, and it seems that nto doing so was the right \n> choice.\n> \n> One of the reasons that the dashed format is being removed is a real \n> technical one, though: git aliases. They never supported the dashed \n> format, since they never were real executables (that's the whole point of \n> an alias, after all).\n> \n> So even with the above path thing, you'll still have to use the spacey \n> version for any aliases you use.\n\nCome to think of it: Maybe commands like \"git pull\" should, when\nspitting out warnings, refer to \"git help pull\" rather than\n\"git-pull(1)\" now. I do like it spacey, but the man issue is confusing.\nIt continues with the links in the help/man pages, of course.\n\nMichael\n"},{"id":"88200","messageId":"7vej4g92io.fsf@gitster.siamese.dyndns.org","threadId":"15138","inReplyTo":"20080822082908.GA29475@isilmar.linta.de","subject":"Re: [kernel.org users] README and ChangeLog files","fromName":"Junio C Hamano","fromEmail":"junio@pobox.com","sentAt":"2008-08-22T21:13:35Z","receivedAt":"2008-08-22T21:13:35Z","isPatch":false,"sender":{"key":"gitster@pobox.com","avatar":"https://avatars.githubusercontent.com/u/54884?v=4"},"body":"Dominik Brodowski <linux@dominikbrodowski.net> writes:\n\n> On Thu, Aug 21, 2008 at 04:20:25PM -0700, Junio C Hamano wrote:\n>> Yes, and the fact that it is not \"git-foo\" that was deprecated, but\n>> running the dashed-form without adjusting PATH was, makes it impractical\n>> to issue warning messages before the transition actually happened.\n>\n> So I can safely ignore the warning in the release notes which states\n>\n> | but users are again strongly encouraged to adjust their\n> | scripts to use \"git xyzzy\" form, as we will stop installing\n> | \"git-xyzzy\" hardlinks for built-in commands in later releases.\n>\n> ?\n\nNo.  What the above means is that the deprecation and removal of git-foo\nfor builtins (such as \"git commit\") are not done in 1.6.0, but it will\neventually happen.  So your two choices are:\n\n (1) stop using dashed form \"git-foo\" and replace them with \"git foo\" form\n     everywhere in your script now and won't worry about this anymore; or\n\n (2) add this to the beginning of your existing scripts:\n\n     PATH=$(git --exec-path):$PATH\n\n     and later do (1) when you have time.\n\nBut if you choose to do the latter, please do so after promising that you\nwon't complain like this time everybody did, saying \"Oh, I knew it was\ncoming but I have postponed doing it for a long time because it continued\nto work\" ;-)\n\n[jc: git list cc'ed]\n\nMaybe we should try to see how hard it would be to issue warnings when\n\"git-foo\" form is used for builtins before we declare that dashed form is\ndeprecated for builtins, and start warning when the deprecation actually\nhappens.  As many people on this thread suggested, it would make the\ntransition easier.\n"},{"id":"88227","messageId":"48AF4740.2030202@garzik.org","threadId":"15138","inReplyTo":"7vej4g92io.fsf@gitster.siamese.dyndns.org","subject":"Re: [kernel.org users] README and ChangeLog files","fromName":"Jeff Garzik","fromEmail":"jeff@garzik.org","sentAt":"2008-08-22T23:09:52Z","receivedAt":"2008-08-22T23:09:52Z","isPatch":false,"sender":{"key":"jeff@garzik.org","avatar":null},"body":"Junio C Hamano wrote:\n> Maybe we should try to see how hard it would be to issue warnings when\n> \"git-foo\" form is used for builtins before we declare that dashed form is\n> deprecated for builtins, and start warning when the deprecation actually\n> happens.  As many people on this thread suggested, it would make the\n> transition easier.\n\n\nI would tend to prefer a compatibility package or similar.\n\nMy fingers have long learned \"git-prefix<tab>\" for several commands. \nAnd Fedora puts so much crap in /usr/bin anyway, I don't see it a big \ndeal to have all those names in the directory.\n\nActually, I bet git-foo, git-bar, and git-blah could all link to the \nsame compatibility script, which simply invokes \"git $command $args...\"\n\nThat would be nice.\n\n\tJeff\n"}]}