Re: Help with UTL_TCP

"Thomas Kyte" <tkyte_at_us.oracle.com> wrote in message

> >The question I have though is why would you want to implement FTP as a
database
> >service? I can see some possibilities (e.g. FTP'ing LOBs), but from an
> >architectural point of view, networking services do _NOT_ IMO belong in a
> >database.

> I rather like the simplicity of having every thing in one place. Do you
believe
> databases should be able to send email? (I do, one of my databases sends
about
> 500-1,000/day).
>
> If email is OK, why not ftp? (hey database, everynight at midnight I'd like
you
> to run this procedure that creates a csv file using utl_file and then ftps it
to
> this other machine that is expecting it). True, we could do this outside the
> database but if you could do it in - why not? The script you develop for
HP/UX
> as a cron job might (or not) work on AIX.
<snipped>

Yes and no. :-) Thomas, I am specifically talking about developing network services (servers) in Oracle.

Yes, I agree that using a database as the repository for an e-mail system is a great idea. Pretty high on my wishlist if for the standard POP clients to provide database support. Even if this has to be via ODBC.

As for driving the processes on the server platform with Oracle job scheduling, is a good idea as this provides for cross-platform deployment, access via a web front-end, Delphi/VB, or even plain NT batch or Unix sh scripting.

No IMO if the idea is to write an "embedded" network server in the database. Like writing a FTP server in using UTL_TCP. Or an e-mail server.

How long have we had SendMail and the various other (mostly OpenSource) servers? And for how long have we suffered the problems with security exploits in these? To re-invent the wheel, redevelop these services using something like UTL_TCP.. I just fail to see why you want to do that. Why you want to risk that. Why you want to plow resources into developing, testing and maintaining that. A robust multithreaded SendMail or FTP server is _not_ a matter of simply slapping a few lines of code together.

Why not _integrate_ the existing services with Oracle instead? Very feasible when dealing with OpenSource.

If the idea is to write embedded network _clients_ in the database, then that is a cool idea. Again this address the issue of portability. There are none of the server side security issues. Also, the client side is much easier to develop and maintain. Heck, if the aim is just to support the basic POP3 protocol, doing a simple e-mail send... a cool idea. (something I would like to see supported natively by more databases)

It is just with the idea of adding foreign network services into Oracle that I personally do not like and fail to see how the benefits of that, outweighs the security risks, the development effort and all that.

--
Billy