Tickets :: Comment on [#13070] Deterministic Asset Names

* Your Email Address
* Spam protection	Enter the letters below: . ..__ . ,. ,.__ \| \|\| \ \./ \./ [ __ \|/\\|\|__/ \| \| [_./
Comment	> Just saw this last comment... > >> but since the assets (e.g. compressed js files) seem to be randomly >> named, when the request for assets goes to an arbitrary backend, they >> will not match the filenames produced by the replying horde instance. > > They are not "randomly" named. The name is the cache key - it is a > hashed combination of the filenames contained in the file + the > modification dates of those filenames. > > Pages that use the same list of js/css files can re-use the existing > files. Modifying one of the constituent files will cause a new > filename to be used/created. > >> of course i could just collect all assets and serve them from a >> separate place (that's probably what i'll do), or as you suggest in >> the conf.xml place the assets directory on a nfs share. but still i >> figured it would be nice if the assets had a deterministic name, then >> this would not be necessary. >> >> additionally it would improve cache efficiency, since random names >> also means that every asset has to be cached once per backend on a >> reverse-proxy. > > Why? The files should be named the same. i.e. the first connection > to imp on a cluster results in a js file being created for the > mailbox page -- all users will then use this file.
Attachment
Watch this ticket