Saving data to stacks via CGI

[Andre Garzia]

On Dec 4, 2005, at 3:38 PM, Alejandro Tejada wrote:

> this is great, but just noticed
> that this option is not included
> or recommended in any message, tutorial
> or page about using cgi in this platform.
>
> Does exist risks for data loss when
> multiple users save data to one stack
> in the cgi folder?
>
> Does exist the real possibility that many
> users try to write the stack at the same
> time and erase other users contributions
> or (worse) damage the stack?


Alejandro,

yes, that is always a risk. This can happen, and in my own paranoic  
opinion, if it can happen, it will happen. I'd recommend one of two  
approaches to solve your problem.

	(1) Use a second CGI/stack as a queue organizer. Your main cgi would  
pass request to change the data stack to this queue organizer that  
would put all requests on a queue and execute them when possible.  
This is some work but it will keep you coding a 100% transcript  
solution that will scale fine. The trick is: received a request then  
queue, you can use a simple text file for queue and the standard file  
locking algorithms for prevent concurrent access to the queue. This  
can even be combined into one single CGI, no matter how many  
instances of the CGI is running, they will all queue requests and  
execute them in order, if you execute the queue after answering the  
browser then you wont clog the user access and your queue routines  
will run on dead time (dead time is the time between your cgi  
answering the browser and it's exit from memory, during that period  
you can do whatever you want since there's no http connection anymore  
and no one waiting for answers, just use that CPU time!)

	(2) That's the easy one. Use a RDBMS as data storage. It's not 100%  
transcript since it involves some knowledge of SQL. You can use  
MySQL, most ISPs support it to one degree or the other, or you can  
opt to go with embedded solutions like altSQLite and Valentina. If  
you use a RDBMS that supports transactions (such as postgreeSQL) then  
you won't even need queue routines.

Cheers and good luck
Andre