From: James F. <jam...@ex...> - 2010-03-31 21:21:13
|
done some further analysis on this and have come to a few 'forks in the road' * Adam Retters vision of /sql/system/sql/{database}/{schemaName}/{tableName} is great ... I would amend this too something simpler like /sql/{database}/{schemaName}/{tableName}/ and this can be the concept of a mounted database from anywhere in collection hierarchy. * this also means we get REST and friends entrypoints * additionally, we may want an alternate configuration e.g. if we consider introducing the concept of a new protocol e.g. rdbms:exist:////sql/{database}/{schemaName}/{tableName}/ which leaves us a bit more wiggle room instead of a mount point in the database e.g. we probably want to make sure people know when an rdbms is being queried in some situations. * from an xpath, xslt and xquery point of view using this protocol should mean nothing special is needed (other then initial setup) to use data from rdbms * I agree with Adam Retters approach of lazy eval though I would like to investigate using an xpath parser approach that generates sql statements ... probably could use a lot of Adam's existing sql extension for the underneath bits * as for registering and configuration I think if its a 'mount' point in the database approach then we can use collection.xconf which could have a snazzy xquery app behind it for CRUD entries. Otherwise the rdbms:exist protocol would be configured 'elsewhere' ok, now reality .... my plate is full (of xproc) and I probably should not get distracted ... so anyone who wants to pick up and run with this tidbit feel free, I have decided to step away from this for the time being. J |