"Wide" UIDs

["Herring, Ed" <ed...@intel.com>]

We are considering going to "Wide" UIDs: 8 digits.
Does Apache have any limitations on UID length (greater than 2 bytes)?

Ed Herring
IT/OPS/AMR2
Engineering Computing
Austin, TX

Desk: 512-314-1133
Cell: 512-917-8480

Re: "Wide" UIDs

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

Herring, Ed wrote:
> We are considering going to “Wide” UIDs: 8 digits.
> 
> Does Apache have any limitations on UID length (greater than 2 bytes)?

There is the implication that UID's are integral.  Provided the apr_uid_t
is correctly defined/detected, and that there aren't too many mismatched
uid_t definitions or uses, this should work with an int64_t.

You can go higher; for Windows, apr_uid_t is a pointer.  It is allocated
out of the pool which required the initialization of the target uid field.
On windows, a uid is a UUID, and apr has the additional comparison calls
for equality to make the developers life simpler.  (Comparing uid's
directly under APR is obviously a nonstarter).

Note that changing uid_t would further require system calls such as
fstat structures to return properly aligned data.  If it was only the
"apr" uid value that you are changing, you may discover some invalid
uid_t declarations that should be apr_uid_t, within apr itself or within
third party applications.

I hope this answers your question.