Notes

HTTP Server

In the current HTTP implementation for a new HTTP server a new PORT number must be used. It is not sufficient to use a new path.

This is particular important for HTTPs servers as every HTTP server is started with a Client CA (certificate authority) directory. The directory holds certificates of trusted certificate authorities.

FROM notes file

processes instead of threads

-> main process does not grow.

... so it could run forever

-> threads tend to grow so dont use threads but processes

-> not multi processing ... no single cpu nor single machine

more

big ania does not need to fork another process so size does not matter. .. so not constraint by size but speed. .. try to not thread in big ania

.. if it starts read last state ... eg it reads a file that contains where we were ... in some structure ... .. big ania files ... big ania should know about what it reads .. eg it can estimate how big it is.

more

on the otherside tasker does not know how big receiving size is ...

more

tasker cannot send a command to big ania ... only say that a command has finsihed.

more

big ania does only manage ...

more

a subshell per tasker. subshell blocking read for exit line of tasker. once it received exit line of the tasker ... it gets the semaphore of the common pipe with ania. then it writes the exit line and takes the semaphore down and exits.

more

all subshell AND big ania

... subshell runs out of memeory, process id or someone kills subshell ... .... what happens? subshell had the semaphore.

when starting subshell ... when it retuns exit code ... if it is a crash ...> ania should check status of the semaphore and probably should take the semaphore done. .... what about the information of the tasker??? so ania should inform big ania about the mistake and the "manager" should decide what to do: eg run the process again or continue or whatever.

more

ania could signal subshell that semaphore is down.

more

technical redirect question

ania must know the exit code of a tasker

shell subshell, subsubshell

( ( cp a b ) && echo blah )

.... echo blah could be something that uses further resources (pipes with ania)

"exed ls"

.... maybe remember who setup the semaphore ...

set -xv; ls .> show commands and parameters to command

... how to debug shell ....

more

regarding one machine with multiple processors vs multiple machines

o big mama makes decisions o taskers run commands

eg tasker_x has job to encrypt.

so for example if the system is under load, big mama knows the load and descides that another machine should exeucte the encrypt task, so instead of calling tasker_x it could call a tasker_y thats only purpose is to send the task to another machines task receiving tasker.

do NOT have a tasker that checks the load and sends the task to a remote machine. becuase what if mama or big mama kills that tasker_x (or it overloads or runs out of memory and kills itself?) ... what about the remote machine? the remote machine is not informed aobut it ...

but if big mama makes the decision it also knows how to compensate a failure or deal with special conditions.