----- Original Message ----- > From: "Robin H. Johnson" <robbat2@xxxxxxxxxx> > To: "Ceph Development" <ceph-devel@xxxxxxxxxxxxxxx> > Cc: "jonathan lacour" <jonathan.lacour@xxxxxxxxxxxxx> > Sent: Monday, June 8, 2015 8:52:21 PM > Subject: rgw static-website hosting design WIP: how S3-like should we be with endpoint types? > > Hi, > > I'm working on implementing S3 Static Website hosting on top of RGW, > based on the initial work that Yehuda did (wip-static-website branch). > > I have a scratchpad here: > https://github.com/robbat2/temp-ceph-rgw-static-website-wiki/wiki > WIP Codebase here: > https://github.com/robbat2/ceph/tree/wip-static-website-robbat2-master > > As part of writing up a full design for it, I'm wondering if we should > try to be "as close to S3 as possible", or try to improve the experience > for the administrator rather. > > In S3, they explicitly add some DNS endpoints that are used for > accessing the bucket in website mode. Via those special endpoints, there > are MANY constraints applied [1], and you pretty much can't use them for > anything else. > > At this point in early implementation, it seems possible that rather > than requiring organizations to have a separate DNS subdomain/wildcard > scope for websites, that we can make the decision based on headers: > If(bucket has website metadata set) { > if(request has auth headers) { > use normal REST mode > } else { > use website mode > } > } > > I've got it mostly working to make this behavior, but I'm wondering > about fallout/breakage, and unintended consequences. The codebase is > pretty ready for plugging the alternative logic with hostnames in as > well. > > The most notable one so far, is that under Amazon's S3, the website > endpoint will server up only fixed pages for directories (no listing); > and the regular REST endpoint will serve listings if the > bucket/directory has a public ACL. > > Another side-effect, and I think that's an improvement, is that if you > use a browser to just access a bucket this way, you'll automatically get > HTML error pages, instead of the XML blobs (as an admin of a RGW install > previously, the non-technical users telling me the rendered HTML from > their errors would have been a huge help instead of them being scared of > the XML). > > [1] The list of S3 website endpoint constraints is here: > http://docs.aws.amazon.com/AmazonS3/latest/dev/WebsiteEndpoints.html > A few options (not necessarily mutual exclusive): - anonymous access gets the website endpoint handling, authenticated gets the REST - read operations returns website, write operations handle REST - configurable (whether gateway handles either, or which) I don't have a strong opinion, more of what users prefer. Yehuda -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
