[Oledb-dev] Connection Pooling

[Shachar Shemesh]

postgresql at bryden.co.za wrote:

>>Do you have any specs on what pooling should look
>>like?
>>    
>>
>If you need to understand connection pooling, here is a link that explains
>connection pooling :http://www.webopedia.com/TERM/C/connection_pool.html
>  
>
I'll retort with a link of my own:
http://m-w.com/cgi-bin/dictionary?va=spec
And, just to make sure you don't miss out on the reference:
http://m-w.com/cgi-bin/dictionary?book=Dictionary&va=specifications&x=0&y=0

The thing most lacking from the link you sent me is the second word 
(first non-small word) in definition 2a of the second link. If *you* can 
deduct what startup parameters to use, when is it ok to reuse a 
connection, whether session parameters need to be reset between sessions 
etc. from the link you sent, I bow my head in humility before your 
superior reading comprehension skills.

>With all due respect, if you are writing a DB driver and you don't
>understand connection pooling, then I am a bit worried
>  
>
Ok, here's "economics of open source 101". You get an OLE DB driver. You 
have to pay for it with neither money nor respect.

The actual program you get varies in quality. Some of it is real junk. 
Some of it is downright brilliant. Most are somewhere in between, as is 
the case with commercial and proprietary software.

Where free software does diverge from proprietary software is when you 
take a program that is in between, and want to move it a little toward 
the former group. Then it is time to pay up.

There are several possible forms of payment. One is actual money. Since 
you already sent me an email in private, and since I have already sent 
you a link explaining the various forms that you can get support, I will 
not elaborate more on this avenue here.

A second form of support is to provide your own. You can certainly sit 
down and write your own connection pooling implementation for PgOleDb. 
You can even send the implementation here, and I promise that if it's 
good enough, I'll integrate it into the main PgOleDb, so that you don't 
have to fork the project just for this feature. If I felt in a nasty 
mood, for example, because someone hinted that I don't know something 
fairly basic merely because they misread a question/request I made, I 
would have referred you to an answer Alan Cox once gave regarding the 
Linux kernel. When asked why didn't Linux have a driver for some webcam, 
his replay was "because you didn't write one".

The sort of form, however, that support does usually take place in free 
software projects is where you ask *nicely* the programmer for a 
feature, and if the programmer has the time, (s)he will do it for you. 
Not for money. Not, even, because we need the feature ourselves, but 
because we care enough about the program we created to want to make it 
perfect.

However, please do bear in mind that when doing that, you do have to 
give an incentive, and maybe even do some of the work. Now, it's obvious 
that you have quicker access to the specifications (we have a linguist 
who specializes in the origins of the English language on the list, so I 
can leave it to him to give you the exact history, but let's just agree 
that the word has some relationship with the word "specifics") of what 
connection pooling should look like *in Ole Db*. I mean, if you don't 
have access to the specs, how do you know whether the same connection 
can be used simultaneously by two threads, whether you need to reset 
session variables after opening a new session, etc. If you wrote a 
program that uses such a feature and didn't have the docs on hand 
explaining the specifics, I'd be a bit worried.

The request I made to you was no different than the request I made to 
the people who want PostGIS support - let me know what PgOleDb needs to 
provide, with specifics (a.k.a. - specifications, also named "specs").

          Shachar

-- 
Shachar Shemesh
Lingnu Open Source Consulting ltd.
http://www.lingnu.com/