Re: [dsfjdssdfsd] Any plans for drafts or discussions on here?

[Francis Dupont <[e-mail]>]

 In your previous mail you wrote:

>    If you take the executable from Solaris and copy it onto your HP-UX
>  machine it will not run.

=> it depends: this is true (i.e., won't run) for an application
written in C/C++, but false (i.e., will run) for an application
written in python or ocaml. So there is another level between
binary and source code compatibilities.
 To come back to the main subject, as an application developer I have
the choice between the OS, the crypto library (e.g., OpenSSL) and
home made to provide a (P)RNG/(D/N)RBG to be used to generate
not predictable (BTW it is nearly always prediction resistance,
not backtracking resistance :-) values (e.g., transaction IDs),
and to generate crypto material (I cite the two because constraints
are very different, for instance for not predictable values
I really want a not blocking solution).

>    My point was that there were not 3 sources (Paul mentioned 3)
>  available to an application, there is just one: the one that's part of
>  OS that the application happened to be compiled for.

=> I should have said "to run over" (BTW with virtualization it is
not so trivial...)

>  > => change your application developer?
>  
>    Ha, ha. Funny. I have considered firing myself several times.
>  
>    The issue was not the code I was writing, it is the annoying habit of
>  linux driver developers to change data structures (the one in question
>  was for an 802.11 device) in gratuitous ways that cause breakage. I
>  have never had that problem with freeBSD. In fact, the inability of an
>  application to just compile from one OS release to another is touted
>  as a feature by some of the more zealous linux people.

=> I'd like to join you as an old BSD man in Linux zealot bashing
(e.g., it is not a gratuitous change: they just never read the previous
code until they get conflicts from multiple definitions with the same
name :-) but we shall be off topic...

Regards

[e-mail]