Re: tcpsvd - Allow only x connections from somebody/hour

> i think this is an interesting problem.  the djb-style suggests something like
> this:
> 
>   tcpserver <host> <port> limitip <service-program>
> 
> where "limitip" is a program i'm trying to dream up.  it's supposed to:
> 
> *  keep a persistent database of TCP connections
> *  get its parameters from environment variables set in tcprules
> *  won't grow its database beyond a configuration time computable size
> *  be simple

 Although I'm a fan of the "one job, one program" approach, I don't
think it's a good idea to separate 'limitip' from 'tcpserver'. I
believe that the 'limitip' functionality should be included into
'tcpserver', just as tcpsvd does. Why ?

 - No persistent data.
limitip needs a persistent database in the filesystem. The database
must be up-to-date at any time. What happens if the machine reboots ?
Stale data. You don't want that. And you don't want a timestamp thing
to check whether the database entry is up-to-date - too complex.

 - Use less resources.
The point of limitip is to limit resource consumption. If you separate
it from tcpserver, you'll fork & exec for every connection, even the
ones that will just be dropped by limitip. You don't want that.
Moreover:
How do you know a connection has been freed and it's OK to accept a new
one from the same IP ? You know because tcpserver's child has died.
limitip needs to know to update its database... so it has to fork & exec
the real server, and stay alive the whole connection time. That means
2 processes by connection instead of one, and another read/write round-
trip for data going through. Ugh.

 Limiting connections by IP is a useful feature, but it's a job for
the super-server. tcpsvd does it well.

 Now, about Maurice's idea : I don't know what would be simpler.
 - Either patch tcpsvd to add a 'max connection per amount of time'
configuration option.
 - Or write a program executed between tcpsvd and its child, that
would update tcpsvd's configuration directory. That wouldn't be as
simple as it seems : it would probably have to be persistent too,
which is a bad idea. (Not even counting the permissions problem if
tcpsvd setuid()'s before exec'ing.)

-- 
 Ska