Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: Solution for 01555
> > Now, I understand, that given the time and resource limitations we all
live
> > under, Oracle maybe couldn't do all this and still charge a reasonable
price
> > for its software. Still, an architecture liable to abstruse errors like
> > "delayed block cleanout" simply sounds incomplete to me.
>
>
> Actually, I think you got this wrong. The rollback read-consistency
> model has been in Oracle since V3. That's about 1983 or thereabouts.
> It's of course been improved and made many times more efficient since
> then, but it's not a haphazard new concept that just made the light
> of day with 9i. No matter how much you may have read about it in
> Oracle's marketing pamphlets.
>
>
> 1555 is NOT a new error. Been there since I can remember working
> with Oracle. The delayed block cleanout as a cause of it is, however,
> a consequence of a new way of handling blocks in cache that came in
> V7. IIRC.
>
>
A couple of comments. I did not mean to imply that Oracle just introduced consistent reads and its associated problems in 9i. Of course, they've been bragging about their architecture for much longer than that. All I'm saying is that, even in its ninth reincarnation, Oracle doesn't protect its users against snapshot errors, which, in my opinion, the RDBMS should prevent internally.
> All this to say: it's not a new concept, subject to polishing.
> It's been thought out very well and it works exceedingly well at
> providing an architecture where writers NEVER EVER block readers,
> something that no other RDBMS in the market at the moment can provide,
> AFAIK. That is good.
Not quite so.
http://carnagepro.com/pub/Docs/Postgres/mvcc.html
http://www.google.com/search?q=multi-version+concurrency
vlad Received on Fri Jan 24 2003 - 02:57:54 CST