[Mono-dev] Mono's future

James Mansion james at mansionfamily.plus.com
Sat Jul 9 06:45:54 EDT 2011


On Wed, 06 Jul 2011 20:56:20 +0100, Rodrigo Kumpera <kumpera at gmail.com>  
wrote:
> Solaris remains a dying unix that it's becoming increasingly irrelevant.

I'm not sure that's a very sensible view, given that a) Oracle have it now
and b) the standards of engineering in it are so different from those in
Linux.  I'm hoping for a resurgence. ;-)

I'm not sure its becoming any more irrelevant if you actually care about
data integrity and support.  I suppose the popularity of 'databases' with
questionable transactional integrity and durability shows I might be in
the minority, which seems unfortunate, but there you go.

Its easy to write off posts like this one:

http://milek.blogspot.com/2010/12/linux-osync-and-write-barriers.html

as being anti-Linux and pro-Solaris - but if you look back over the various
discussions on (say) the Postgres mail lists regarding data persistence
semantics, and you recall the fuss over the apparent slow performance on
MacOS X in DB benchmarks when they enabled the disk cache flushing, and
you hunt around for discussions of the write barrier reliability in various
Linux filesystems (and DM/MD etc), its apparent that the issues can't be
dismissed as shallow bugs, but design issues - or in fact problems with
the process in terms of who actually owns 'make sure the data is on the
platter'.  There have been problems for a decade or more in terms of
getting Linux to really persist data, and I'm not convinced they will
go away any time soon.

James

(And the relevance is: please don't take Linux-centric design decisions
and assume its somebody else's problems to do the port - like the
X and Gnome/XFCE people have.  POSIX was supposed to avoid this, and
the old we-are-better-than-windows-because-we-follow-open-standards
approach seems to have gone by the wayside as soon as the market share
supports doing so.)


More information about the Mono-devel-list mailing list