i went down to this presentation at westminster university on amazon web services. it was completely non-technical. at least code wise. it was surprisingly interesting. they have about 6 semi-different web services that use either soap or rest interfaces. according to barr, it’s split with about 80% using rest. good. can we please kill off soap before it wastes anymore time? the most interesting would have to be the mechanical turk and amazon s3, simple storage service. the most novel use of the amazon catalog would have to be this music association flash thing called liveplasma.
the storage system really implements the kind of system i would like for storing files. it removes you from any file system oriented limits. however, do appear to be limited to a 5g per file size. the docs seem fairly straight forward. one odd thing would appear to be the creation of their own authentication mechanism. at least they’re using the authorization header. it reminds me of the google file system that was talked about a few years ago. distributed and redundant. i wouldn’t store anything private in it that wasn’t encrypted though.
someone should really be selling something like the storage system. by nodes and drop them in with no configuration required. access everything using get and put over http. million dollar idea. the only problem being that it could probably be implemented in a few months or less. many people would probably get it wrong though by making it too complicated or tying you into some proprietary crap.
overall, positive experience. i guess the only difference between amazon and google is that amazon’s core business is selling goods and google’s is advertising. kind of makes you long for the day when corporations couldn’t just go into multiple sectors. definitely would help the microsoft situation.
update: i forgot to mention barr’s flippant response to a comment from the audience about amazon’s stupid one click patent. barr mentioned that amazon wanted to reach out to developers and build a community. someone in the front row mentioned that maybe they shouldn’t do things like patent one click if they wanted to encourage small developers. barr’s laughing response was that it’s business and they have to do it so that they can cross-license with other big companies if someone tries to sue them. the man in front responded that that doesn’t really do anything for the little guy. barr side-stepped the issue and changed the subject. further proof that software patents do exactly zero to help the little guy like they’re claimed to do.
I would think that webDAV, for which implementations are available, could met this need. I believe there is also a specification. I used clients that work well, like any file explorer GUI. Of course, the same issue with all solutions is that different implementations can’t seem to meet the specifications the same way. What do you think it takes to get accurate implementations?
webdav is similar, but adds a lot of unnecessary features for the target usage. the spec is too complicated and also requires xml. xml jammed into headers no less. the one feature from webdav that may be useful is a way to get a directory listing in a data format. however, i suspect the target usage is a scenario where the files stored are known by the client application that directly use the urls. it may be possible to do directory listings. i didn’t really look at the docs that much.