On 27/12/2008, at 2:41 PM, Chris Anderson wrote:
This does nothing to address the restrictions on the character set of database and design doc names. Antony's patch looks like it works as advertised. If people are happy with the way it reorganizes couchdb data files, then I don't see a reason not to apply it (especially with the addition of a slug).
It was built before the changes to the view server, so it doesn't apply cleanly to head. If the change is approved in principle, then I'm happy to do the work to make it slug-based i.e. <filtered-name> + <md5>, where filtering removes all the special characters, prefixes and sequences, and doesn't append an md5 if the filtered and raw names are the same.
I'm ambivalent about the fact that it nests the derived data (e.g. views) within the database directory. I think there's a valid argument against that, based on the physical separation of the canonical data and the derived data, but as I say, I'm ambivalent. I'll change it, if you want, to use two directories per database, one for canonical data and one for derived data.
Antony Blakey ------------- CTO, Linkuistics Pty Ltd Ph: 0438 840 787 The difference between ordinary and extraordinary is that little extra.
