Given the new proxy deployed for Task cluster (and bug 1162126), do we still need proxxy? The double-proxy seems awkward, possibly introducing more places for things to fail. If not, can we describe what the value is of having both proxies in contrast with the problems that generates for failure detection, retries, etc.
Probably not for TC artifacts. We still need it for stuff coming from tooltool/pypi/ftp/etc. I think
We certainly need proxxy for inside SCL3 (for everything) otherwise we could very easily adapt https://github.com/taskcluster/s3-copy-proxy to work for everything else... This seems ideal to me as we (my team) now that we are multi-region are going to put a pretty high priority on optimizing this as data-transfer is one of our biggest spends if it does not work correctly ;) Currently I wrote the proxy to only accept one source (which happens to be an s3 bucket) we could easily change this to work like this: <proxy host>/ftp <proxy host>/tc <proxy host>/pypi Each of those sub urls (or hostnames) would map to a different source.