Hi,
A very good initiative! I will certainly use it in my scheme-osa project.
Do you envision setting some sort of Snow Patrol to assure consistency, portability and quality of the Snow packages? Is there any grand library design concept in Snowfort, or is it intended to be just a loose system of individual packages?
I have no idea how the self-control of package authors work in practice in other repositories of this sort. But I wonder how Egg or PlaneT deal with duplication of effort (for example, 10 HTTP oriented packages, rather than one or two), or proliferation of low quality "hello world" entries.
Jan
P.S. To illustrate why I wonder about the latter I point to a great democratic initiative for publishing scientific papers -- arxiv, the repository of e-prints in Physics, Mathematics, Computer Science and Quantitative Biology, http://arxiv.org/.
The idea is great and the arxiv archive has served us well for many years. However, there are some problems with this repository, as David Bacon, the owner of Quantum Pontiff blog (http://dabacon.org/pontiff/?p=1418), explains:
"... while the arxiv is a amazing tool, one of the problems was that the volume of papers was high and, to put it bluntly, the quality of these papers was not necessarily so great."
Consequently, he established the scirate.com site to rate the papers (initially quant-phys), in order to filter out the noise. More details in the article I cited, which is sort of FAQ on subject of rating, presented rather humorously.
Afficher les réponses par date
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 2-Mar-07, at 11:49 AM, Jan Skibinski wrote:
Hi,
A very good initiative! I will certainly use it in my scheme-osa project.
Do you envision setting some sort of Snow Patrol to assure consistency, portability and quality of the Snow packages? Is there any grand library design concept in Snowfort, or is it intended to be just a loose system of individual packages?
You raise a very interesting issue. We all hope that the packages will be of high quality but the model set forth by Snow allows anyone to submit any package (the minor requirement is that the snowball must be structured like a Snow package but other than that anything goes). This freedom of speech is really important so that no one feels excluded. I personally vow not to interfere with the submission of packages except to deal with profanity and vandalism. I am also willing to share the responsibility of maintaining the snowfort with others to avoid dictatorship.
I can think of a few ways to control quality:
1) Have some way to test portability of packages. The package should have self tests. If these self tests pass on N host Scheme systems, then that package gets a grade of "N". Now who should write the self tests? Perhaps the users should write the tests so that the package author cannot trick the system. The tests could be moderated by the snowfort maintainers to make sure they are fair (or maybe the package can be deleted by the snowfort maintainers if users complain that the tests written by the author are insufficient).
2) Count the number of times a package is downloaded. This is not a very good measure because you need to download it once to install it, and you won't download it again whether you use it on a regular basis, or never. Perhaps "snowman" should have a command for removing the local installation of a package, and that command could contact the snowfort to decrement the download counter. I hope that doesn't count as a violation of privacy...
3) Have some kind of voting system so that users can indicate their level of appreciation for a package.
4) The snowfort could keep a list of user comments for every package. This could be combined with the voting system (in fact maybe you can only vote if you write at least a minimal comment).
I have no idea if any of this makes sense in practice. Also, I don't have anytime soon to implement this. Any takers?
Marc
snow-users-list@iro.umontreal.ca