[GRASSLIST:9538] Re: Rasters not displaying from old (Windows) Gr ass Projects

[Glynn Clements]

Chris George wrote:

> Thanks for your detailed responses.  I should say, first, that I am 
> enormously impressed by the GRASS software and by the efforts of the 
> people who continuously maintain it and provide so much help to other 
> users.
> 
> What started my comments were the remarks from someone else to the 
> effect that one might as well use the latest cvs version as the latest 
> stable one.  I expected that a "conservative" user should be advised to 
>   use the latest stable release.  Your remarks have confirmed this, thanks.
> 
> Concerning another message in this thread, no, I have no evidence that 
> the CVS version contains more bugs.  I don't use the CVS version, so I 
> could collect no such evidence.  But I am only talking about degrees of 
> risk.  Most open-source projects adopt the same strategy of "stable" and 
> "development" versions and will accept that the development one is more 
> likely to contain bugs, especially if one means so far undetected ones. 
>   In absolute terms it might contains less, in that a discovered bug has 
> been fixed in the development version but not yet backported.  But in 
> general use of the development version is more risky.

A few things to bear in mind regarding the nature of bugs and
debugging:

1. The size of the GRASS code base is huge. By some measures, 5.0.x
was on a par with XFree86 (6.x has had a lot of unmaintained code
removed), but both the user base and developer base are much smaller. 
Bugs can lurk for years without being discovered (I've found a couple
of bugs in 6.x which have existed since 4.x, and quite a few which
were introduced in the 4.x -> 5.0 transition).

2. A lot of the heaviest or most demanding users use the CVS mainline
because of the additional features, so it potentially gets more
thorough testing than the stable releases.

3. Bugs are often introduced in bursts of active development then
removed on a more continuous basis. For any given module, the version
from just before a substantial update may be the most reliable.

-- 
Glynn Clements <glynn at gclements.plus.com>