# Re: Mercurial vs Updated git HOWTO for kernel hackers 3 messages from 2005-06-28 to 2005-06-29. Participants: Horst von Brand, Christopher Li, Kyle Moffett. Thread: https://gitlist.dev/t/1040 ## Christopher Li, 2005-06-28 18:47 Subject: Re: Mercurial vs Updated git HOWTO for kernel hackers Message-ID: <20050628184755.GA2255@64m.dyndns.org> URL: https://gitlist.dev/e/20050628184755.GA2255%4064m.dyndns.org In-Reply-To: <200506282154.j5SLsETL010486@laptop11.inf.utfsm.cl> ``` On Tue, Jun 28, 2005 at 05:54:14PM -0400, Horst von Brand wrote: > > And is exactly the wrong way around. Even RCS stored the _last_ version and > differences to earlier ones (you'll normally want the last one (or > something near), and so occasionally having to reconstruct earlier ones by > going back isn't a big deal; having to build up the current version by > starting from /dev/null and applying each and every patch that ever touched > the file each time is expensive given enough history, besides that any Mercurial store a full text node when it detect the delta gets too long to reach certain point. So what you describe here will not happen on mercurial. Having it append only is a very nice feature. > error in the file is guaranteed to destroy the current version, not > (hopefully) just making old versions unavailable). It also means that > losing old history (what you'll want to do once in a while, e.g. forget > everything before 2.8) is simple: Chop off at the right point. You can still chop of the history before the full node, but rebuilding the repositories. Mercurial save some much space that you would wonder why do you what to chop the history if you can keep it. Chris ``` ## Horst von Brand, 2005-06-28 21:54 Subject: Re: Mercurial vs Updated git HOWTO for kernel hackers Message-ID: <200506282154.j5SLsETL010486@laptop11.inf.utfsm.cl> URL: https://gitlist.dev/e/200506282154.j5SLsETL010486%40laptop11.inf.utfsm.cl In-Reply-To: ``` Andrew Thompson wrote: > Petr Baudis wrote: > >>Mercurial's undo is taking a snapshot of all the changed file's repo > >>file length at every commit or pull. It just truncate the file to > >>original size and undo is done. > > "Trunactes"? That sounds very wrong... you mean replace with old > > version? Anyway, what if the file has same length? It just doesn't make > > much sense to me. > I believe this works because the files stored in a binary format that > appends new changesets onto the end. Thus, truncating the new stuff > from the end effectively removes the commit. And is exactly the wrong way around. Even RCS stored the _last_ version and differences to earlier ones (you'll normally want the last one (or something near), and so occasionally having to reconstruct earlier ones by going back isn't a big deal; having to build up the current version by starting from /dev/null and applying each and every patch that ever touched the file each time is expensive given enough history, besides that any error in the file is guaranteed to destroy the current version, not (hopefully) just making old versions unavailable). It also means that losing old history (what you'll want to do once in a while, e.g. forget everything before 2.8) is simple: Chop off at the right point. -- Dr. Horst H. von Brand User #22616 counter.li.org Departamento de Informatica Fono: +56 32 654431 Universidad Tecnica Federico Santa Maria +56 32 654239 Casilla 110-V, Valparaiso, Chile Fax: +56 32 797513 ``` ## Kyle Moffett, 2005-06-29 00:12 Subject: Re: Mercurial vs Updated git HOWTO for kernel hackers Message-ID: URL: https://gitlist.dev/e/A595C063-7F6E-49AD-A797-1133C731E7E3%40mac.com In-Reply-To: <200506282154.j5SLsETL010486@laptop11.inf.utfsm.cl> ``` On Jun 28, 2005, at 17:54:14, Horst von Brand wrote: > Andrew Thompson wrote: >> I believe this works because the files stored in a binary format that >> appends new changesets onto the end. Thus, truncating the new stuff >> from the end effectively removes the commit. > > And is exactly the wrong way around. Even RCS stored the _last_ > version and > differences to earlier ones (you'll normally want the last one (or > something near), and so occasionally having to reconstruct earlier > ones by > going back isn't a big deal; having to build up the current version by > starting from /dev/null and applying each and every patch that ever > touched > the file each time is expensive given enough history, besides that any > error in the file is guaranteed to destroy the current version, not > (hopefully) just making old versions unavailable). It also means that > losing old history (what you'll want to do once in a while, e.g. > forget > everything before 2.8) is simple: Chop off at the right point. If we have versions A through A+N, Mercurial will create a new revlog file and store a new full version when the total size of the changes between A and A+N is greater than a certain amount, effectively ensuring that retrieving the latest version of a file is O(size-of-file) instead of O(size-of- file*revisions). This is the same speed as RCS for the tip, and significantly faster than RCS for non-tip, which is crucial for merges. Cheers, Kyle Moffett -- There are two ways of constructing a software design. One way is to make it so simple that there are obviously no deficiencies. And the other way is to make it so complicated that there are no obvious deficiencies. -- C.A.R. Hoare ```