Thread

Hmm, should ACL[] be toastable?

Tom Lane <tgl@sss.pgh.pa.us> — 2000-07-29T19:08:39Z

I notice that ACL lists are represented as arrays of aclitem, which
means they are now theoretically toastable.  (In practice, I haven't
finished fixing all the routines that touch ACLs, but will soon.)

Do we need long lists of ACLs?  If so, is there any danger in giving
pg_class a toast relation?  It's a tad closer to the heart of the
system than pg_rewrite, so I'm not quite sure if that will work or
not.  Jan?

			regards, tom lane

Re: Hmm, should ACL[] be toastable?

Jan Wieck <janwieck@t-online.de> — 2000-07-29T23:23:02Z

Tom Lane wrote:
> I notice that ACL lists are represented as arrays of aclitem, which
> means they are now theoretically toastable.  (In practice, I haven't
> finished fixing all the routines that touch ACLs, but will soon.)
>
> Do we need long lists of ACLs?  If so, is there any danger in giving
> pg_class a toast relation?  It's a tad closer to the heart of the
> system than pg_rewrite, so I'm not quite sure if that will work or
> not.  Jan?

    In theory it should work, in practice, I don't know.

    Since pg_class is really close to the heart of the system, it
    is created a little different during bootstrap. This  causes,
    that  setting  relacl  to  storage  'x' doesn't automatically
    create  a  toast  relation  for  it  during  bootstrap.   And
    therefore,  the  toaster  should  only try to compress, never
    move out (to where?).

    Someone could later create such  a  toast  table  with  ALTER
    TABLE  ...   if  he wants to give it a try. And we could warn
    him not to do so before we really stress tested it.

    Is that a compromise?


Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck@Yahoo.com #