[[meta copyright="Copyright © 2009 Free Software Foundation, Inc."]] [[meta license="""[[toggle id="license" text="GFDL 1.2+"]][[toggleable id="license" text="Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled [[GNU_Free_Documentation_License|/fdl]]."]]"""]] [[meta title="Have Debian's apt use ftpfs"]] \#hurd [[IRC]] channel, 2009-03-13. By the way: I had another GSoC idea: (a) make ftpfs / httpfs unsable enough for... (b) have Debian's apt / whetever not use a direct ftp / http connection, but instead use /ftp/ftp.debian.org/whatever etc. (c) Make unionfs unsable enough, so that it can be used together with nfs (or whatever) to sarve as a cache for /ftp/. uhm... you mean all of this as a single GSoC project? indeed I have the intention of turning ftpfs/httpfs into something that can be used by other tools instead of a library but that requires them to grow completely new interfaces, making them suitable for interactive use you are right though that this would be a nice GSoC project... no idea why I never thought of it in this context I think I proposed this idea already some months ago on some mailing list. I don't think it's too much work. Or why do you think? it is considerable work it requires implementing a non-trivial, directory based FS interface (c), OK, but why (a) and (b)? also, I'd probably reimplement them on top of libcurl... no need to maintain our own wheels :-) That for sure, yes. But for starterst, we can just take ftpfs as it is. I think I explained the issues in some past discussion, though I don't remember where it was And just teach apt to open /ftp/... instead of calling whatever function to retrieve a ftp file. you want progress tracking for one you want the client to see redirects (OK, that might not be necessary for apt-get... it would be necessary for other use cases like a web browser, though) you want proper error handlin http://thread.gmane.org/gmane.os.hurd.bugs/16575/focus=16795 none of this can be expressed by opening a single virtual file it would require adding a completely new interface, either using custom RPCs, or a directory with several virtual files for various aspects of the interaction (I prefer the latter variant, though I guess some people would argue in favour of the former :-) ) anyways, I'll add it to the list OK, I see. Nevertheless, don't you think that a more simple implementation of my idea (also leaving out (c) for now) wouldn't be a suitable GSoC project? apt-get is probably quite a good example use case for this, so you are probably right that it would be useful to develop this together (I'll leave out (c) though -- I would consider it a distinct project. we can place a followup task once (a)+(b) is done :-) ) Yes, sure.