Re: Is Ingres yet relational?

Paul Mason <Paul_member_at_newsguy.com> wrote in message news:<c9mnmk01gq_at_drn.newsguy.com>...
> In article <bcb8c360.0406020907.635f2818_at_posting.google.com>, Tony Douglas
> says...
> >
> >Paul Mason <Paul_member_at_newsguy.com> wrote in message
> >news:<c94b9l01ll4_at_drn.newsguy.com>...
> >

<snip>

> I agree it should be easier. The fact that it's not is, I suspect, the result of
> a chicken-and-egg situation where not enough people use UDTs to request changes
> and UDTs have not been perceived as easy enough for many people to try to use
> them.
>
> My semi-tongue-in-cheek remark (you'll note the smiley) was in part to point out
> that they're not as hard to implement as many people suppose.
>

Sorry - I think I leapt on that harder than it probably merited :) It's a thing Roy H and I have been batting about for a while. I dunno - maybe it's a bit like J2EE where there's a lot of boiler plate that you can have generated for you automatically rather than have to go through a lot of "mechanical" stuff over and over. I kind of lost the will to live after a bit of piddling about with it.

> Yes and no. Because the code you write for UDTs is in effect put into the 'guts
> of the DBMS' - alongside the code to deal with built-in complex data-types (i.e.
> non-machine types) such as date, decimal etc.
>

And I suppose that's what I don't like about it so much - I'd like UDTs to work (to the DBA or appropriate person for defining data types) at a higher level than that. Obviously at the end of the day UDTs will worm down to this level, but not defined at that level.

> Look, I'm not suggesting that the current situation with Ingres UDTs is ideal
> but I just found it amusing that in the same thread where people are blithely
> talking about stripping SQL out of Ingres, we also have people balking at some
> much more straightforward coding. I realise these aren't necessarily the same
> people - but nonetheless it tickled me - hence the comment and the smiley.

Indeed - I think it highly unlikely that a bedroom hacker will contribute anything of great value to the core Ingres code line - there's just too much of it, and I remain to be convinced that the open source community has the bottle to do the type and level of testing required to work on an enterprise level DBMS. MySQL is a joke, and PostgreSQL is great at all the things programmers would like a DBMS to be great at, but not so hot at the things a DBA would like a DBMS to be good at.

A lurking genius who could integrate a type system on the lines of Haskell or SML into ADF/DMF and make it sit nicely next to QUEL would be fab though ! :))))))

• Tony