Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.misc -> Re: rman 10G

Re: rman 10G

From: Vladimir M. Zakharychev <vladimir.zakharychev_at_gmail.com>
Date: 19 May 2006 00:18:56 -0700
Message-ID: <1148023136.576094.161530@j73g2000cwa.googlegroups.com>


>However I strongly disagree with your assertion on Oracle not being
>downwards compatible. In fact it has always been.
>And then again: has Mickeysoft ever been downwards compatible? Oracle
>had to release patches when Mickeysoft released NT4 sp3, replacing
>Winsock1 with Winsock2 !!

And do you say that upgrade to Winsock2 was a wrong move for Microsoft and their customers didn't benefit? Well, quite a few apps had to be updated to make use of it, but MS provided backwards compatibility layer that emulated Winsock1 on Winsock2 to make the transition as smooth as possible. And Oracle is releasing patches regularly anyway, to fix their own defects and to implement workarounds for defects or make use of new features, introduced in operating systems, so I don't really think it was much of issue.

Actually, I see backwards compatibility requirement one of the main obstacles to making better software in general, Microsoft suffering even more than others: they know they did it wrong, but they can't just say "Hey, we did it the wrong way, let's just scrap it and build a new system from scratch that will (hopefully) do it right." But there are tons of applications built on that wrong design, using its quirks and tailored to live with its limitations. You can't just tell all vendors to scrap those apps and redo them the right way, too. So you have to keep all those quirks, limitations, bugs and "features" for the sake of backwards compatibility. Software evolves, but if it is to be backwards compatible it has to inherit from previous releases, and not always the right things. Remember the problems caused by more restrictive security model in XP SP2? When they did things right, just partially, it immediately broken tons of apps that were designed for, and exploiting, the wrong model.

Pretty much the same with Oracle - they fix a bug, or introduce a new access path, or change costing model of some particular access path in the CBO (actually making it better and more accurate) and all of a sudden you need to re-tune your SQL because it was tuned for old CBO behavior and new behavior causes unanticipated side effects. They upgrade PL/SQL and you need to fix your old code that exploited quirks in the old engine (yes, it's usually latent bugs in your code that you need to fix, but it was working correctly on older releases and now the compiler simply barks at you. And if that code is wrapped and you don't have sources, you're in trouble.)

Generally speaking, yes, Oracle was always backwards compatible within documented limits (I mean, it doesn't have to be, and usually isn't, compatible with desupported releases,) but to a degree. There's always "to a degree" when we talk about backwards compatibility, because the only way to make software fully backwards compatible is to stop evolving it.

Best regards,

   Vladimir M. Zakharychev
   N-Networks, makers of Dynamic PSP(tm)    http://www.dynamicpsp.com Received on Fri May 19 2006 - 02:18:56 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US