diff options
author | Sascha Silbe <sascha-pgp@silbe.org> | 2010-07-21 17:30:43 (GMT) |
---|---|---|
committer | Sascha Silbe <sascha-pgp@silbe.org> | 2010-07-21 17:30:43 (GMT) |
commit | 5ab8ddb8ddb8dc6925cb3bb5f52e0a06b7a1c439 (patch) | |
tree | 01e6ca6d9d6eaa54052480200f039c8e50f0166f /docs/instances-vs-programs | |
parent | d4fe57f377d8f78f02a9e5f6d9c201b4e9eb86f4 (diff) |
rearrange source tree to match that of the tarball
Diffstat (limited to 'docs/instances-vs-programs')
-rw-r--r-- | docs/instances-vs-programs | 34 |
1 files changed, 34 insertions, 0 deletions
diff --git a/docs/instances-vs-programs b/docs/instances-vs-programs new file mode 100644 index 0000000..3f18b58 --- /dev/null +++ b/docs/instances-vs-programs @@ -0,0 +1,34 @@ + +== Problem: Containerize Activities or Instances? == #none + + We're presently on the fence about whether to isolate different instances of + activities from one another. + +Situation: + + Ivan doesn't think it's worth it; Noah is concerned about controlling the + spread of viral documents. + + Michael doesn't feel strongly about it; he doesn't see much difference in + implementation difficulty. + + Michael thinks the usability issue is going to come down to the fact that we + rate-limit by container. If instances are packed into one container, then + long-running background downloads could really mess up your browsing + experience. If they're allocated per-instance, though, we don't (at the + moment) have any hard bounds on allowable resource usage. + + Either way, if the user is going to rely on Rainbow (as opposed to the app + itself) to do rate-limiting, then we might find ourselves constantly + twiddling limits in order to make a decent browsing experience. + + Also, do we ever need to firewall instances, or just activities as a whole? + + (Incidentally, how *are* we going to rate-limit activity startup?) + +Plan: + + + +Followup: + |