RE: [Fwd: Re: installing squeak on Debian Woody]
Ian Piumarta <[email protected]> Tue, 2 Sep 2003 03:01:23 +0200 (MET DST)
| Newsgroups | gmane.comp.lang.smalltalk.squeak.foundation |
|---|---|
| Message-ID | <[email protected]> |
Hi Andreas, > What we were trying to achieve with the download page was more along the > lines of having a single page with ALL relevant information and direct links > to the stuff you need to run Squeak as well as being easily editable. Check > out our example at > > http://minnow.cc.gatech.edu/squeak/3262 That's a _lot_ better. A total pain to update whenever new versions arrive (and too wide to format into a reasonably-dimensioned browser ;), but _much_ better all the same. Now, I have no substantive issues with any reasonable download organisation (with one very big exception -- which you'll be able to figure out, later in this message). So what follows is kind of taking my position to an extreme, but I'm going to indulge in it anyway as a kind of thought experiment... Rather than having nrows*ncols of VM dowloads for Unix on that page (with so many version numbers screwed into everything that I certainly wouldn't want to spend an hour editing it and testing every link each time a new release comes out), explain why having that huge Unix download matrix is better than having a single link to www-sor.inria.fr/~piumarta/squeak where the entire matrix is reproduced (almost) verbatim (except that you missed a few archives -- there should be 32 of them in total ;) _and_ is generated automatically (it takes me less time to rebuild the entire download matrix than it does to open the editor to change the one variable defining the version number ;) _and_ in which every one of those download links is verified automatically each time I rebuild the page? I don't know if anyone has noticed this but the process is entirely automated -- from build to publishing the new downloads (and almost all of the re-sync with SF too ;). If I change something for Unix and bump the version number, I run one script and 5 minutes later all affected archives amongst those 32 have been rebuilt -- and the links to them on the download page updated and verified for correctness. > which should give you an idea about what we're after (note that the "info" > link on Unix VMs goes straight into your primary web site for support) I still can't see the advantage of reproducing the download links in some place that's only indirectly connected to their content's origin: (1) the matrix is incomplete; (2) it's unmaintainable (without inordinate amounts of work each time it changes -- and that's sometimes every few hours for Unix during beta test); and (3) I often pay little or no attention to problem reports that arise from VMs packaged by other people -- and I suspect you'd feel similarly about repackaged win32 VMs that had been randomly scrambled. (I recall Stef, I think, although I could be wrong, getting quite upset [in public -- which I was not happy about] over problems with Solaris that arose from somebody repackaging stuff.) > The whole point of this download page is to make clear what the "standard > home" for Squeak downloads is, be able to update it easily and provide > up-to-date information on the individual ports. As far as Unix goes, the most up-to-date you could possibly get would be to remove the entire Unix matrix and replace it with one link to the original Unix download page. Or even better, one link and a list of the platforms for which archives are available via that link. (I would be more than willing to keep that list up to date -- my OS versions change far less frequently than do my VM versions. ;) Apart from anything else, when I'm not distracted with other things, new betas come out every few days -- sometimes every few hours [*] -- and nightly automatic builds of the live source tree are available for download too. The "download" section for new Unix Squeak users on my page is also much more detailed, and includes a lot of information that is specific to Unix. That section is not going to go away, and reproducing a small "generic" subset of it elsewhere (along with a bunch of archives that merely incites people to dive in potentially underprepared) is simply bad for new Unix users. Thought experiment terminated. ;) Cheers, Ian [*] Just a silly example: I released beta9 two hours ago. I'm about to release beta10 within the next hour -- to incorporate your new sqOpenGLRenderer fixes (unless I fall asleep first ;). If it wasn't for the total automation of the process, which makes creating a new release (and updating its download links and verifying them) trivial to manage, beta10 would not be happening any time soon. It seems strange to want to deprive anyone of that kind of "liveness" by keeping a second-hand set of links to stale archives...