[Freeciv-Dev] Re: Server from Client

[Raimar Falke <hawk@xxxxxxxxxxxxxxxxxxxxxxx>]

On Fri, Dec 28, 2001 at 12:29:22PM -0500, Daniel L Speyer wrote:
> On Fri, 28 Dec 2001, Per I. Mathisen wrote:
> 
> > On Fri, 28 Dec 2001, Daniel L Speyer wrote:
> > > Unfortunately, it doesn't.  An extended version of it might.
> > >
> > > I recently tried it for checking when a user had connected, but it was all
> > > the same '012' as tons of other messages.  I haven't looked at what would
> > > be needed to expand rfcstyle to what's needed.
> > 
> > The server often doesn't use the correct C_* attributes to messages. I'm
> > not fixing this in the console cleanup I am writing now, but I'll have a
> > go at it once that is done if none else wants to.
> > 
> > > What's needed is really something of a mess.  As I see it, it is nessesary
> > > to use C's string handling and raw unix io to both find the reply to our
> > > question (with potentially anything having been printed since we
> > > asked) and then to parse it in spite of numerous free-form strings.
> > 
> > No. With rfcstyle, you send the server a command, and then wait for a line
> > that starts with "100" or one of the warning/error codes. Once I get the
> > cleanup patch done, this number will be followed byseveral quoted
> > arguments, like this
> > 
> >     100 "Per I M" "AI"
> > 
> > which isn't a big problem to parse.
> > 
> > One problem might be that string results like "AI" might be translated...
> > there are a few such problem cases in the code... I am not sure what is
> > the best way to deal with it, but I think I would rewrite those few cases
> > to ensure rfcstyle sends untranslated information, or perhaps by exporting
> > console_rfcstyle and make if ... else tests for it in those cases.
> > Definitely solvable.
> > 
> > >  I'm
> > > starting to seriously think that every query should have a unique flag
> > > that the server echos back with the answer.  Does anybody have experience
> > > with this?
> > 
> > Wouldn't that be like rfcstyle except every answer has a unique number? I
> > don't see the point.
> 
> OK, here's the problem:
> Client 1 sends message to server
> Client 2 sends message to server
> Server answers client 1's message
> Client 2 picks up answer to Client 1's message
> Creash, burn, error, end of game
> 
> Now, messages on the network connections are somewhat direc ted, but it is
> quite possible to cause a message to be sent to another client, so parsing
> becomes a real mess again.
> 
> Also, consider:
> Client sends message 1
> Server answers message 1
> Client sends message 2
> Client listens for responce to message 2
> Client pick up responce to message 1
> crash, burn, end of game
> Server answers message 2
> 
> Now, all commands could be required to pick up any responce, but this
> looks like trouble to me.  It's sort of anti-encapsulatory and seems to be
> begging for future errors.  Plus, it means that adding a responce to any
> server command would mean massively re-working client code.
> 
> Putting unique tags on all messages could avoid all this without
> sacrificing extensibility.

Although I don't know what rfcstyle (the undocumented command) does in
detail I know that we have such unique identifier for every command:
the request-id. So in rfcstyle or another new style the server could
prefix every output with the client-connection and the request-id
which caused this output. This makes easy and always correct matching
possible.

        Raimar

-- 
 email: rf13@xxxxxxxxxxxxxxxxx
 "Make it idiot-proof and someone will make a better idiot."