Re: ora-1450 even though max key length is well below the max

Anurag Varma wrote:
> Maxim Demenko wrote:
> > Anurag Varma schrieb:
> >
> > > That said, 3118 limit seems to be for a 4K block size. Are you sure
> > > that your block size is 8K?
> > >
> >
> > >
> > > Anurag
> > >
> >
> > I have just verified it on a 9.2.0.6 / Linux and got 3218 ( not 3118 )
> > for 8k blocksize , on 10.2.0.1 / Linux /8k blocksize - same results as
> > yours (i.e. 6398) , in the Note Chuck mentioned there are limits for 8i
> > which correspond to mine. It seems, the limit is version/plattform
> > dependent.
> >
> > Best regards
> >
> > Maxim
>
> Yes, good point. I knew it was version dependent ... however, did not
> realize it
> would vary so much in platforms.
> In solaris 9.2.0.6 I get the limit as 6398 (don't remember exact figure
> now).
>
> Anurag

I think the issue is more complicated than reproducing with a script with various versions. Look at the bugs for the error. Note there are bugs for things like function based indices, db's with multiple blocksizes, alter index online (blowing the IOT journalling table used), MV's over dblinks and so forth. I think what we can abstract from this is that lots of the relevant code has been futzed with and dragged in different directions over the point versions as well as the major versions, so the limit can vary from 1/3 db default blocksize on up. There are probably more bugs we don't even see on metalink, I'd guess that 9205 having a higher limit is one of them. Just because it doesn't throw an error on creation doesn't mean there won't be problems down the road...

jg

--
@home.com is bogus.
"Hooray class of '06. All these years have payed of (sic)."