Hi Martin

On 09.04.2010 12:15, Martin Bartosch wrote:
> Determining the installed software version at runtime can be done with 
> something similar to
> 
> perl -e 'use OpenXPKI::VERSION; print $OpenXPKI::VERSION::VERSION'
> 
> but this only works if OpenXPKI was built properly. I found that sometimes 
> (depending on build process and packaging) this information may 
> not be accurate. We could add an API function to get 
> this version number, of course.
> 
> Reporting the package version is a different thing, it's highly 
> dependent on the target system distribution, and I don't think we can 
> find a canonical way to determine package versions regardless of system 
> type. It will not work on systems where OpenXPKI was not installed 
> via a package manager, as well.

Yes. Seems that this "about" problem should be the last in the todo
list, but should be there anyway.

Spent quite some time reading websvn. Seems that the idea to roll
back last year is unsatisfactory. Suggest to forget it.

Freebsd project is rapidly moving everything to openssl-1.
Today it has openssl-0.9.7 or openssl-0.9.8 installed as a part of
the base system (different branches of freebsd has different
versions of openssl here). Instead of the base openssl, a user can
choose to install openssl-1 from ports (ie from package) on any
freebsd branch. Hence openssl-0.9.8 is not available at all in some
branches. (Assume today nobody installs anything by hand.) My
attempt to persuade freebsd people  to keep openssl-0.9.8 as a
separate port failed, despite the fact that many other ports are
reported been incompatible with openssl-1. openssl-1 deliberatly
neglects backward compatibility with openssl-0.9.8. That is why I
had to declare a freebsd collection of openxpki related ports as
"broken" for certain branches of freebsd, and "incompatible" with
many modern installations which have openssl-1 installed.

As of the OpenXPKI, the problems related to openssl-1 are as follows:

- naive checks of openssl version which assume version to begin with
a string "0.9."

- implicit assumption about low-level data types declared in
openssl, which have changed in openssl-1. This is not related to
cryptography.

- calls to low-level functions that vanished or changed in
openssl-1. This is not related to cryptography.

- different approach to deal with cryptography, which is very
similar to our cryptography abstraction concept.

So let us answer questions about our future tactics. Let us forget
about non-western cryptography for a while.

1. Shall we just migrate to openssl-1 abandoning support for
openssl-0.9.8, or shall we support both versions of openssl? Support
of both versions simultaneously just in a xs code could be very
hard, because openssl heavily uses preprocessor stage redefinition
of data types, with which it is not easy to cope when programming in
xs instead of plain c.

2. Shall we try to prepare a new cryptographic backend called
openssl-1, in addition to existing backend called openssl-0.9.8?
This would justify our old claim

http://www.openxpki.org/docs/crypto-plug.html

that we are ready to add new separate cryptographic layers, and will
retain support for openssl-0.9.8 in a more beautiful fashion.

This approach will require some code fixing to return to
cryptographic modularity concept though. And could buy us extra
bonus returning the non western cryptography a sort of automatically.

Regards,
Sergei


------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
OpenXPKI-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/openxpki-devel

Reply via email to