Open stdweird opened 8 years ago
@NZJourneyMan @ajf8 comment?
I think I may have added (or at least changed) that code... if so, it wasn't intentional.
@ajf8 may remember more when he's back from holiday but ISTR 4896589 was due to the code taking too long to failover between dead proxies because curl's timeout doesn't work as you might expect and most of the time you want head_timeout not timeout. It's possible that for consistency the same fix should be moved to the non-proxy code path, but AFAIK a "fast-failure" is less useful there because there's only one server anyway.
ok, that makes sense. but a head timeout is still usefull if there are a lot of things to download, even if it's not from proxies. the current refactoring uses the same timeout for both proxies and non-proxies (it's easier and cleaner that way). so unless there's a good reason not do it like that, i'd rather leave it like that.
what is (currently) not addressed is the (lack of) per file head_timeout.
currently ncm-download uses a per file timeout with check for age via HEAD only if proxies exist. if not proxies exist, a component wide timeout is used and no HEAD checking is performed. is this intentional?