Jon Rascal wrote:
Hi,
I want to create a virtual file system where metadata is stored in an LDAP repository. Metadata includes owner, mod-time, create-time, access-control-lists, pysical file path, and, of course, LDAP's hierarchy instead of a file-system directory base.
Files are to be stored on physical medium and (preferably pyton) code used to bridge the request (over https), the metadata, and the physical file stream.
Users and Keys are also stored in the LDAP directory.
Is this a bad use for LDAP?
Seems like it should be OK, but it certainly won't be great for performance. Though as you're basing this in python, I guess that's not one of your top priorities.
As a matter of fact, I wrote a couple fileservers before coming onto the OpenLDAP Project; I wrote an NFS server for IBM mainframes back in 1988, and I wrote an AppleShare server for Unix in 1995-96. (Which by the way was also the world's fastest, as OpenLDAP is today the world's fastest LDAP server.) So from my perspective, ya seen one hierarchical data store, ya seen 'em all.
General purpose data retrieval protocols like LDAP are at a slight disadvantage because they're designed to work for anything, including arbitrary-sized requests and data. Purpose-designed filesystem protocols can use fixed-size requests and data because the full scope and nature of requests is known in advance. That gives dedicated protocols an advantage as far as parsing and encoding performance. But with that said, OpenLDAP's parser is pretty efficient, so the potential difference is small.
I'm also looking for an LDAP guru to consult for this particular project.
JR