IPC, RPC, Protocol to use between fe/be?

Simon Perreault nomis80 at videotron.ca
Fri Jun 15 21:37:07 PDT 2001


On Thursday 14 June 2001 20:40, Gerard Beekmans wrote:
> Ok XML again. Simon for one doesn't like the notion of XML and perhaps
> there are others who think the same. I don't blame them, part of me is
> asking the same questions if there aren't alternatives to XML. It's a
> good thing we check that out, in case others ask these questions again,
> we at least can say "No, absolutely not. XML is by far the best way to
> go, because of this, this and that. Oh and of course becuase of these".

I just don't want to go with XML because of its coolness aura. I want solid 
proof that we can't choose anything better than XML.

> Like my other email said: if you really wanted to go with shell
> scripting, some RPC implementation could just dump a script instead of
> an XML profile and have it do the same. Though, I do see limitations and
> problems because of bash' nature being a shell, not a program that runs,
> can run in background, can poll for waiting messages on a socket, etc.
> Bash wuld need to start some helper program which in turn needs to be
> able to SIG<something> bash to have it to do something alternate (say on
> SIGALRM bash goes check for a file /profile-useme and work on that one.

Create perl daemon. Have it listen on port X. Authenticate, authorize, then 
take input and use a lot of system()'s. Send feedback as needed. Bash isn't 
the central element. It's like the worker slave.

> I don't really know how we would deal with bash in such a way, I just
> see a worl of complications showing up that we won't have with non-bash
> (then still remains the question of XML or some custom ALFS-message
> format. Ok, why re-invent the wheel when XML suites already. We'd most
> likely make up our own ML version (ALFML? It's all ML, the only thing
> that makes an ML HTML, SGML or XML is the DTD behind it).

Why an ML? What do we gain? Sure, we know where each command fits in context, 
and we can have hierarchies and inheritance, and we can enforce data 
conformity, but is it any use?

The data we have is contextless, it is meaningless, and it is procedural. It 
is read once and used once, always in the same order. We don't need to 
enforce data conformity as a shell script (which an XML profile would just 
end up being) can not conform to anything, it is freeform.

> Enough mumbling, main thing I wouldn't mind learning about are
> XML-alternatives, if any. I'd like to be able to answer:
> 	"No, absolutely not. XML is by far the best way to go, because
> 	of this, this and that. Oh and of course becuase of these".
> Right now, I couldn't do that without using very weak reasons.

Maybe we should not look for alternatives, but we should ask ourselves, what 
do we need? Our data is procedural and read once/use once. I suggest shell 
scripts. As the data, of course, not as the program that will use them.

-- 
Simon Perreault -- Public key: http://nomis80.linuxfromscratch.org/nomis80.gpg
-- 
Unsubscribe: send email to alfs-discuss-request at linuxfromscratch.org
and put unsubscribe in the subject header of the message



More information about the alfs-discuss mailing list