On 2/26/16 1:30 AM, Tom Lane wrote:
> As the patch is presented, I agree with Peter that it does not really
> need a format number bump. The question that has to be answered is
> whether this solution is good enough? You could not trust it for
> automated processing of tags --- it's easy to think of cases in which the
> schema/object name separation would be ambiguous. So is the tag really
> "strictly for human consumption"? I'm not sure about that.
Well what are those tags for? They are not used by pg_restore, so they
are for users. My understanding is that the tags help in editing a TOC
list for use by pg_restore. What pg_restore actually reads are the
OIDs, but the tags are there so users can edit the files. The tags can
also be used for ad hoc automatic processing. They are not sufficiently
delimited and escaped for robustness in all cases, but it can be done if
you control the inputs and know what to expect. But this is the same
problem before and after my patch.
Both of these cases are helped by my patch, and both of these cases were
pretty broken (for the object classes in question) before my patch.