Steve (and all), It's been a while, and with my old brain I'm trying to remember what had gone on to date on this subject. This is at least partially an attempt to "remember out loud" (I'll try not to explicitly note everything as I remember or find it). ;-)
The last recommendation (by me, but based on words that Jeff Licquia supplied earlier, before I became aware of the other XDG specs mentioned below,) reads like this: <quote> See also: * XDG Base Directory Specification (ref1) * xdg-user-dirs (ref2) * enum GUserDirectory and subsequent sections in the GLib Reference Manual: Miscellaneous Utility Functions (ref3) Together these set a number of conventions intended to organize user specific files. " where ref1 = http://standards.freedesktop.org/basedir-spec/basedir-spec-latest.html ref2 = http://www.freedesktop.org/wiki/Software/xdg-user-dirs ref3 = http://developer.gnome.org/glib/unstable/glib-Miscellaneous-Utility-Functions.html#GUserDirectory </quote> Ok, I should have gone back a step further--the original words, copied (by me) almost exactly from words suggested by Jeff in an email, iirc, included this line: "The XDG Base Directory specification is recommended, but not mandated." Your objection to that was: "I object to saying that the FHS "recommends" the XDG spec. https://lists.linux-foundation.org/pipermail/fhs-discuss/2011-May/000171.html"; I (iirc), in a (perhaps misguided) attempt to address your concern--(after Jeff commented that we should remove the recommendation) simply deleted that particular line. I'm guessing that your concern (at least partially), looking at it now, is that the simple "See also" sort of implies a recommendation by the FHS? (Just for the record, although I'm sure you sense it, I would really prefer to see it be a recommendation. But, also for the record, my deletion of the line calling it a recommendation was an honest but, in retrospect, shortsighted attempt to comply with what I took to be "direction" from Jeff.) Anyhow... I would still like to get some mention of those XDG specs into the FHS. At the very least it: * gives users a sense of what could be done to possibly make file handling for them more convenient. (I have let my ~ directory become the jungle--I let my distro (Debian) do pretty much what it wants there, except that I've moved all my "real user data files" out of ~ and into a separate /<username> directory. * lets users know that there has been some thought given to a standard way of doing that (as described in the referenced XDG specifications), so that if they decide to "push" for changes in that direction (in the distro or applications they use) they have a target to suggest to the developers of their distro or applications * lets developers know the same thing, so that if they decide to migrate their distro or their application in that direction, they are aware that a standard exists and can head in that direction So, I'm going to assume you don't strongly object to at least mentioning those specs in the FHS, somehow. So, then I'd suggest wording something like the following: <quote> Some users have expressed a desire to separate the data normally in the users ~ directory into multiple directories, not necessarily under ~. Some thought has been given to that, specifically by the XDG. The FHS is neither recommending for or against such a change, but provides pointers to those specifications for anyone interested. See: * XDG Base Directory Specification (ref1) * xdg-user-dirs (ref2) * enum GUserDirectory and subsequent sections in the GLib Reference Manual: Miscellaneous Utility Functions (ref3) Together these set a number of conventions intended to organize user specific files. " where ref1 = http://standards.freedesktop.org/basedir-spec/basedir-spec-latest.html ref2 = http://www.freedesktop.org/wiki/Software/xdg-user-dirs ref3 = http://developer.gnome.org/glib/unstable/glib-Miscellaneous-Utility-Functions.html#GUserDirectory </quote> If your concern is that there is no plan with respect to migration concerns, let's talk about that. I see that as either a non- or fairly long term issue. See what I said earlier (below--with some additions) about an envisioned migration path. On Sunday 26 June 2011 02:50:23 am Steve Langasek wrote: > On Tue, May 17, 2011 at 08:52:44AM -0400, Randy Kramer wrote: > > I agree that the XDG Base Specification has little if anything with > > respect to addressing migration concerns. > > > > Well, except that I can envision a migration path--I don't see > > anything that precludes a developer from starting to follow the XDG > > Base Specification such that, sometime down the road, when enough > > applications support it, it can be adopted. I mean, the first step > > seems to be the developers specifying things using the symbolic > > names / locations (like $XDG_DATA_HOME) instead of specific > > locations (like /home/<username>). Adding: Distros can initially point all of those symbolic names to the current specific locations (i.e., all to /home/<username> somewhere). At some point in time, when enough applications are using the symbolic names instead of specific locations, distros (or users) can start to point those symbolic names to their preferred locations. > > I guess I would ask you, what does having this as a recommendation > > do to you (or Debian) that makes you want to object to it? I'm > > guessing that it may force you (or Debian) to do something (else, > > why would you object)? > > The concern is not that Debian will be forced to do something; my > concern is that, by having the FHS, which is a widely recognized > standard, recommend the XDG spec, this will be used to persuade > upstreams of various applications already in use to start to follow > that spec without regard to the migration issues, resulting in a poor > experience for users. > > In general, the *only* parts of the XDG base spec which I think it's > reasonable for existing software to adopt are $XDG_CACHE_HOME and > $XDG_RUNTIME_DIR, because these relate to disposable data and there's > a distinct benefit to switching (namely, making it simpler to exclude > these files from a backup policy). For the rest of the spec, a > migration is more trouble than it's worth, and the FHS should take > pains to avoid implying that such a migration is a good idea. If you want wording that recommends $XDG_CACHE_HOME and $XDG_RUNTIME_DIR, but nothing else, I would still like to mention the rest of the XDG specs without recommending them, something like the wording I recommended above. I would add, there is a distinct advantage to me to have my "real user data" separate from all the detritus in ~ so that I can back that up safely and separately. Randy Kramer _______________________________________________ fhs-discuss mailing list [email protected] https://lists.linux-foundation.org/mailman/listinfo/fhs-discuss
