376136 - extend nsIRDFRemoteDataSource Flush() and FlushTo() to take permissions parameter for the file we write to disk

Status: RESOLVED INCOMPLETE

(Core Graveyard :: RDF, enhancement)

Description

[(not reading, please use [e-mail] instead)]

extend nsIRDFRemoteDataSource Flush() and FlushTo() to take permissions parameter for the file we write to disk

dveditz writes:  "It would be nice if the RDF consumer could specify the permissions for different files/uses, wouldn't it?"

One way we could do this would be to add a param to Flush() and FlushTo().

another approach that would not require a change to the interface is to specificy it as a query string on the uri we pass to init() or flushTo(), like file://foo/bar/cheese/blah.rdf?perms=0600

if not specified, we'd default the perms to 0666 so this would not affect any existing code (like in seamonkey, see bug #375102 for details)

Comment 1

[Robert Kaiser]

Actually, it would probably be best if we could default to using the user's umask as already discussed in the other bug.

Comment 2

[Axel Hecht]

Is this a real bug or an RFE? Like, do we have an existing caller that needs this?

Comment 3

[Christian :Biesinger (don't email me, ping me on IRC)]

You can't avoid using a user's umask... (well you could if you called chmod manually)

Comment 4

[Daniel Veditz [:dveditz]]

I doubt "flush" is the right place to specify the permissions. If the file is new then the consumer probably wanted to specify the permissions when they created/opened it.

If the file already exists IMO we should preserve the existing file permissions (we have an nsIFile pointing right at it in flush, "if file.exists newperm = file.permissions" (converted to all the ugly C++ error checking)).