Re: [HACKERS] LZTEXT for rule plan stings

At 03:02 AM 2/26/00 +0200, Hannu Krosing wrote:
>Don Baccus wrote:
>>
>>
>> I still like the idea of "text" being implemented under the hood
>> as lzText for a quick 7.1 release if that idea works out ...
>
>But without TOAST it would result in _undefined_ max tuple length,
>which is probably not desirable.

Boy, I'd sure find it desirable. There's nothing to stop people from
using varchar(8000) or whatever if they want a predictable top limit.
Text is not a standard type, and this wouldn't break standard semantics.

lzText wasn't removed because folks thought it was useless, IIRC,
it was removed because TOAST was an exciting and much more powerful
approach and no one wanted to introduce a new type doomed to disappear
after a single release cycle.

With TOAST, from the user's point of view you'll still have an
_undefined_ max tuple length - the max will just be really, really
large. Sure, the tuples will actually be fixed but large varying
types can be split off into a series of tuples in the TOASTer
oven, so to speak. So I guess I have difficulty understanding
your argument.

If text were implemented as lzText for a quick 7.1, which apparently
was Jan's spin on the idea, then for 7.1 we'd say:

"maximum number of characters you can store in a column of type
text varies"

and after TOAST we'd say:

"maximum number of characters you can store in a column of type
text varies"

Right? The only difference is that the _undefined_ maximum in
the non-TOAST case is "thousands of characters" and in the TOAST
case "gigabytes of characters" but undefined is undefined in my
book.

- Don Baccus, Portland OR <dhogaza(at)pacifier(dot)com>
Nature photos, on-line guides, Pacific Northwest
Rare Bird Alert Service and other goodies at
http://donb.photo.net.