Top things to see in first ALFS
zhouhui at wam.umd.edu
Wed Feb 2 07:43:41 PST 2005
On Wed, Feb 02, 2005 at 08:18:57AM -0700, Kevin P. Fleming wrote:
>Hui Zhou wrote:
>>Jamie, you missed the point of his. The server don't need parse the
>>stream since no profile stream is being sent. Try re read his post.
>>He is voicing another Client/Server model.
>I don't believe this model will work out very well...
>For example, in chapter 6 of the LFS book, we enter a chroot environment
>and run a number of package installations. This really does _need_ to be
>run as a single process if at all possible, the same way nALFS currently
>does it. Jumping in and out of the chroot environment for every single
>command in chapter 6 would be very slow, difficult and likely to cause
>problems that would not occur under normal operation.
The bot has to be at least two processes, one listening on the
network, the other does the job. Between those two, they may talk on a
FIFO or unix socket or simply temp files.
The listening process don't need to be changed through out the whole
process, but it needs track the bot process incase it forks. On
detection of fork, it tracks the fork pid, so it knows which process
to talk to.
In this single command mode, we have to request the bot don't exit on
completing the task, instead, send a signal or message to the
listening process and go to sleep. When new command arives, it wakes
up and work again.
I don't see the necessity ot jumping in and jumping out.
More information about the alfs-discuss