Thomas Weber wrote: > On Fri, Jul 20, 2012 at 11:18:35AM +0200, Philip Nienhuis wrote: >> Admittedly there's some overhead inside. What's the problem with that? >> >> On this box I have a complete MinGW/MSYS development environment >> installed, plus 5 or 6 Octave-MinGW versions including MSYS + a lot >> more. It takes up a bit of disk space, true, but these days that >> shouldn't be a problem. > > I suggest you take a moment and take SF's perspective. >> From their website, they say that they have had 4 millions downloads > *today*. Let's pretend that just 1% of those downloads included msys, > which has about 3 MB according to > http://sourceforge.net/projects/mingw/files/MSYS/Base/msys-core/msys-1.0.11/ > > That's 120 GB of network traffic on one(!) day just for MSYS. > > If you were to distribute this via Amazon's S3 (0.12$/GB)[1], that's > roughly 12$/day. At the end of a year, that's 4000$ - again, just for MSYS. > > FWIW, one of the reasons for dropping the m68k architecture from Debian > was the fact that a full Debian mirror exceeded 100GB in harddisk space > and most mirror operators could not justify that much space. > > Now, the numbers may vary (especially the 1% above is probably too > high), but then again we have only talked about msys and nothing else. > So please, don't take your local hard disk as reference for what forms a > problem in software distribution. > > [1] http://aws.amazon.com/de/s3/pricing/
Thanks Thomas for an enlightening post. Indeed, there are more perspectives and interests than just Octave users, developers etc. But I have ambiguous feelings on this subject. While I can easily perceive that for a single project based on volunteers like Debian the distribution costs issue is very real, I think the SourceForge case is a bit different. Just expanding on your figures: IIRC Octave-MinGW_3.2.4 had around 1 million downloads in total. I can't dig up those for the later MinGW-installers (it looks like the numbers have been reset, perhaps by shuffling around files), but AFAICR they hovered around some tens of thousands. Just assume that all MinGW installers together had 1.5 million d/lds. Imagining a rough average size of Octave's MinGW binaries of 150 MB, download costs of all MinGW binaries from SF in history, at $0.12/GB, would amount to some $27,000 (IMO still an inflated guess). Perhaps I'm spoiled, but <shrug> - I'm not so impressed by that figure. I can't imagine that Sourceforge, whose core biz *is* SW distribution, doesn't have a business model that let them cope with this at ease. If they couldn't they would already have signaled this to the OF site maintainers a long time ago. Until then it simply isn't our problem. More to the point: as long as SF doesn't complain, we shouldn't need to worry too much. Nevertheless I think it is *very* good that you draw attention to some hidden, less favorable and especially, non-durable aspects of "free" software. It would be nice to have an incremental mode for distribution of (Windows) installers to conserve bandwidth. But it should be easy enough to keep on attracting non-devs (if even n00bs) to download Octave. The MSVC version already comes this way, at the cost of another scarce resource (= the packager's (Michael's) time). So I guess my POV is this: If the choice is between * Ease of installation for novice users, implying less frustration with them and better reputation for Octave, and less installation support time for the devs around here, -versus- * Avoiding hidden distribution costs that we yet haven't heard about anyway, I'd simply prefer the former. But admittedly there's a slight itchy feeling. Philip ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Octave-dev mailing list Octave-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/octave-dev
