[Mono-list] Crypto Unit tests on linux

[Piers Haken]

Well, go ahead, but it still seems brain-dead to be reimplmenting
something that's already done. The space argument is fatuous - have you
ever looked at the memory profile of a running mono app?

Isn't it better to get this stuff running in the most portable way
without duplicating whole chunks of effort just to satisfy some far-off
1% scenario?

Piers.

-----Original Message-----
From: Sebastien Pouliot [mailto:spouliot@videotron.ca] 
Sent: Wednesday, October 16, 2002 3:31 PM
To: mono-list
Cc: Paolo Molaro; Piers Haken
Subject: Re: [Mono-list] Crypto Unit tests on linux


Thanks for all three options.

> 1) write a small xml parser that handles just the stuff you need in 
> corlib
> 2) (better, IMHO) write an internalcall that uses the GMarkup parser
> in glib and fills in the RSA and DSA structures.

> > 3) use: Assembly.Load, Activator.CreateInstance & MethodInfo.Invoke

I would find the reflection stuff much easier but Paolo has a good
point. I've participated in the design and implementation of a hardware
token which included crypto, smart cards and biometry (and I even looked
at how we could get a JVM into that box). I know that I would have hated
any design that didn't minimize dependencies as much as possible (as
such design are often very limited on memory space).

However I would like very much, if possible, that the crypto classes to
be implemented using 100% managed code and I know that some other
classes are gonna require a lightweigth XML parser (probably
CryptoConfig but I've not finished looking at the System.Configuration
classes).

So I will put the small parser on my todo list (unless someone else is
interested :-) and comment the current "bad" code so the tests (and
classes) can be compiled back into mono.

Thanks,

Sebastien Pouliot
Security Architect, Motus Technologies, http://www.motus.com/
work: spouliot@motus.com
home: spouliot@videotron.ca