Man, these tables drive me crazy. Is there a good reason I'm missing that we don't it like this?
+------------------+ +---------------+ | page | | page_version | +------------------+ +---------------+ | page_id | ----> | version_id | | current_version | ----/ | page_id | +------------------+ | timestamp | | user | | ... | | data | +---------------+
That way, all your page versions are in one place -- a lot easier to manage than two tables.
Just some woolgathering as I fight my way through some RFEs...
~ESP
On Dec 1, 2003, at 17:43, Evan Prodromou wrote:
Man, these tables drive me crazy. Is there a good reason I'm missing that we don't it like this?
+------------------+ +---------------+ | page | | page_version | +------------------+ +---------------+ | page_id | ----> | version_id | | current_version | ----/ | page_id | +------------------+ | timestamp | | user | | ... | | data | +---------------+
I believe I've already advocated this for the next major revision of the software. (We can transform 'old' into the revisions table and keep the present old_id numbers intact, then add the present 'cur' data after it.)
But first, we branch a stable release of the current version. We _are_ in feature freeze, supposedly.
-- brion vibber (brion @ pobox.com)
wikitech-l@lists.wikimedia.org