[EMAIL PROTECTED] (Cathleen Wang) writes: > -=-=-=-=-=- > > Meeting Notes from Monday 4/30/01 performance meeting. > Comments welcome. > > > * resolving host too often, is there a DNS cache? (gordon) > > http://bugzilla.mozilla.org/show_bug.cgi?id=30917 > > STATUS: phil and selmer noticed just by clicking back/forward on pages > over slow modem they're seeing app "resolving host blah...." phil was > also concerned for people going through proxy servers, we don't need to > do DNS lookup every time, since proxy server address remains the same. > > gagan said what people are seeing is not necessarily reflecting that's > what's happening in netlib (details in the bug). gagan was also > concerned about doing DNS cache and round robin would break servers that > want to randomly assign IP addresses to distribute server loads. > > Decision: > gordon to implement DNS cache and have global pref to control lifetime > (value was set to 900sec in 4.x). No UI for this pref, but can be set > by embedder. > Well, round-robin DNS shouldn't have any problems with this, because the addresses will still be doled out in approximately the same proportions, even if they're not fetched from the DNS server as often. It could potentially, however, hurt schemes that are smarter than round-robin (ie that hand out IP addresses specifically based on server loads or other non-random criteria). As long as the cache time is defaulted reasonably low (and can be turned off by setting cache time to 0), it's probably a win, though a few people are likely to scream. I guess I would vote for a somewhat shorter default (300 seconds?). Ideally, it will someday be possible it will be possible to determine DNS record time-to-live values in an XP way without writing our own resolver. Ah, to dream. :-) Dan --
