> > Or simply call text-BLOBs "textblob" of something like that.
> > What does SQL-92 say about BLOBs anyway?
>
> Nothing afaik. That is why you get different meanings and usages between
> database
> products. I'd like to keep "text" as a useful string type. Conventionally,
> generic
> blobs are just binary objects with not much backend support (e.g. no
> useful
> operators other than perhaps "=").
>
> Imo generic blobs make more sense in a system without the capability to
> add types;
> perhaps a solution for Postgres would look a little different. At the
> moment, the
> frontend/backend protocol is different for large objects and everything
> else, so
> it would be difficult to transparently introduce blobs which behave
> identically to
> types which fit within a normal tuple.
>
> - Tom
Yup, that all sounds very plausible. But, since the meaning diverges between
DB Systems
I would suggest to maybe not enforce text for now (at least not in system
tables).
It has almost the same behavior as varchar (does it ?), and since varchar is
very good now :-)
I would enforce the use of varchar where it fits (like passwd in pg_shadow,
but not prosrc in pg_proc where text is appropriate).
Maybe just to keep the doors open for a larger text datatype in the future.
Andreas