dmd 1.061 and 2.046 release

Nick Sabalausky a at a.a
Sun May 16 11:58:27 PDT 2010


"BCS" <none at anon.com> wrote in message 
news:a6268ff13be78ccc28aed90fd98 at news.digitalmars.com...
> Hello Nick,
>
>> "BCS" <none at anon.com> wrote in message
>> news:a6268ff13bd08ccc2716440002a at news.digitalmars.com...
>>
>>> The same holds for every file in /usr/bin, I wonder what that says
>>> about all the other people who put stuff there. Similar thought hold
>>> for the other bits and places.
>>>
>> Maybe it's my windows upbringing, but I've never liked the idea of
>> having each of my apps spread all across the whole filesystem.
>>
>
> There is something to be said for that, but at least with Linux it's 
> *only* the filesystem that it gets spread across (registry).
>
> I think this is a case where the phrase "when in Rome" is a good starting 
> point.
>

True enough.

Actually, this is something I've often given thought to. The basic problem, 
really, is inherent limitations of hierarchies. There are apps, and then 
apps can have executables, helper executables, asset files, help files, 
settings files, plugins, etc. This is really a 2D matrix with "App" on one 
axis and "Type of data" on the other. So to put it into a hierarchical data 
system (ie, any modern filesystem) one must arbitrarily choose one of the 
axes to be the most significant. Unix traditionally chooses "type of data". 
Windows and the modern OSX package system choose "app" (with notable 
exceptions - registry, user settings directories). My own personal 
preference is "app", but there are certainly reasonable points to be made 
for either approach.

This also gets into why I was a bit disappointed that MS's WinFS project 
died out. I hadn't thought much about it prior to all the talk of WinFS, but 
things like that and iTunes convinced me rather quickly that hierarchical 
filesystems are a bit antiquated for modern needs, and that there are 
definite benefits to be gained from a relational approach even if it's 
nothing more than a system-wide layer on top of a traditional hierarchical 
system (hell, DBMS's abandoned hierarchies in favor of relational long ago, 
and for good reason). But of course, actually pulling that off on a 
technical level, and doing it well, is probably another matter entirely, at 
least if MS's experience is any indication.




More information about the Digitalmars-d-announce mailing list