[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lemonade] Removing attributes, was Re: Taking the knife to METADATA


Cyrus Daboo wrote:


Hi Randall,
--On March 14, 2008 7:17:47 AM -0700 Randall Gellens <randy@xxxxxxxxxxxx> wrote: 


We were trying to maintain some level of similarity with ANNOTATE, and
also provide the client with a way to see the size before fetching
something.
Well ANNOTATE is only experimental and I'm not sure anyone (except perhaps Ken @ CMU) has done anything with it. So limiting ourselves to ANNOTATE syntax should not be a requirement IMHO.
As far as the size goes, we could easily add an option to GETMETADATA that would allow a client to limit the size of data returned and add an option to the response to indicate when a truncation has occurred, e.g.: 

          C: a GETMETADATA (MAXSIZE 1024) /comment
          S: * METADATA (TRUNCATED 2048) /comment "My comment ...."
          S: a OK GETMETADATA complete
I can live with that, as long as this is a part of METADATA/METADATA-SERVER and not yet another extension.
So in this example the client asks the server to return no more than 1024 octets. The /comment value returned is truncated at 1024 octets, and the server also tells the client that the actual size of /comment is 2048.
I think this is more efficient if you make the assumption that for the most part annotations are going to be small. i.e. the client doesn't have to do two roundtrips to get an annotation (one to check the size, the other to get the data). 


That is a good point.
Instead it just does one with MAXSIZE and then, only if there was truncation, can do another get if it really want to.