Re: WANTED : the sqlnet protocol
Date: 1996/04/03
Message-ID: <4jueq1$5ol_at_access1.digex.net>#1/1
In article <31613779.41C6_at_ensco.com>, Stan Driggs <stan_at_ensco.com> wrote:
>Hugo Toledo, Jr. wrote:
>>
>> mark rostron wrote:
>> >
>> > Can anyone please advise if the sqlnet protocol is published anywhere? mr
>>
...
>
>So, can anyone please advise if the sqlnet protocol is published anywhere?
>
Would be interesting to see, but I suspect it is proprietary, so
if you did code up a client oracle would not support you. Not
a pretty thought...
>IMHO, it would be easier to write your own client/server protocol, with the server
>using dynamic SQL or OPI to process requests for data. Has anyone had any experience
>with this?
>
We did this. Actually several times, one with an in-house string-based protocol (so even stupid visual basic programs could talk to our servers), and another with an RPC interface. Both using a unix server process which did the queries using PRO*C Dynamic SQL method 4. Both worked fine. Did in/out variables and everything. Only one cursor though.
The hard parts were:
- getting dynamic sql working. Find ALL the documentation and read the example programs.
- Stupid TCP/IP stacks/software on the windows 3.1 clients. PC's crash, connections hang, GPFs and all that.
The drawbacks are that you need one server process on unix for each client. Uses up a lot of oracle licences that way. You could also multiplex the requests, but then you loose transaction commit/rollback. Then again, if everything were packaged up in stored procedures...
You could also do a RPC implementation of OCI. But I think you would end up doing more network transactions than the other way around. Then you would need RPC on the clients. This can be done in house, or you can buy it. Microsoft supplies RPC on NT, but it is, of course, incompatable with ONC (normal, sun, free) RPC. Sigh.
Cary O'Brien
cobrien_at_access.digex.net
Received on Wed Apr 03 1996 - 00:00:00 CEST