hi ben... what you talk about is almost done: - all files are downloaded and put into the local package-cache however: - MC always queries the server, even if the files exist locally - MC has to decode the complete version history even if you just want to see the commit message... fixes: - I made a local hack to make sure MC always queries the local repo first - when downloading versions don't re-serialize them again when saving to the local package-cache directly write out the loaded bytestream - Using a shared package-cache avoids downloading multiple times the same version - Maybe we should add a separate entry for an mzc which contains an isolated version info - Maybe we should just move to git :) (my premature version browses commits instantaneously) With the first two hacks I got nice performance when browsing history / commit messages. By having a shared package-cache I could improve the load-time roughly by a factor 5... best cami On 2012-04-20, at 15:46, Ben Coman wrote:
This is a second idea for an enhancement to the Monticello Repository browser that at least myself and perhaps others would find useful.
In the Monticello Repository browser, when a package has many new versions and I am wanting to review all of the comments for each package (both for curiosity and also trying to select which I want to load), it is time consuming to do this, having to wait for the selected version to download before seeing the comment, then select the next one and wait, then select the next one and wait, etc...
It would be nice if all comments for the various versions could be downloaded in one go at the start - for example from a single file. I see two different ways this might be done. 1. Monticello clients downloads such information from the repository, updates it and then uploads it upon a Save or Copy. 2. A smart backend server process culls the information from the uploaded mcz files to build a single file.
However when I say 'single file', when the incremental comments build to some maximum size, that incremental file could be archived to a static version which clients would keep hold of, and roll over to a fresh 'incremental' file)
I hope that makes sense, cheers -ben
btw, I didn't chime in before but congratulations with the 1.4 release. Its a bigger effort than I realized and I'm impressed. I'm looking for to using Pharo even more.