I notice that one of the features mentioned in the proposed Adlib base
profile is "the SRU protocol, i.e., HTTP GET/POST CGI requests".
This reminds me of a discussion I had with Adam and some other Index
Data people a few days ago: is it possible to POST an SRU request
instead of GETting it? It seems to me that it _should_ be, but there
are pragmatic reason why it might not be the best thing.
What saith the scriptures? Not very much as it turns out:
SRU, the Search and Retrieve URL Service, is a companion
service to SRW, the Search and Retrieve Web Service. Its
primary difference is its access mechanism: SRU is a simple
HTTP GET form of the service. The request is encoded within
the URL, much like OpenURL, in contrast to SRW where an XML
document is sent to the server.
That's all. And yet there seems no compelling reason why I should get
allowed to do:
GET /services/l5r?operation=searchRetrieve&version=1.1&query=sword&maximumRecords=10 HTTP/1.1
POST /services/l5r HTTP/1.1
Why is this important? Because of the real, imagined and
implementation-dependent limits on URL length. Les Wibberley's famous
search-terms that are 10,000-node chemical formulae are not going to
fare well under SRU-GET.
/o ) \/ Mike Taylor <[log in to unmask]> http://www.miketaylor.org.uk
)_v__/\ "I think it should either be unrestricted garnishing, or a
single Olympic standard mayonaisse" -- Monty Python.
Listen to free demos of soundtrack music for film, TV and radio