[chuck-users] cast dur to time?

[Kassen]

On 07/01/2008, pibsid at suomi24.fi <pibsid at suomi24.fi> wrote:
<snip>

That should be enough.


Yes, what we have is "enough" in a way, we can do all that was suggested
might need to be done (including referring to the birth of the VM hours
after it started) but in order to do it we need to take some routes I find
counter-intuitive on a syntax level.


Requesting time casting sounds to me of requesting different kinds of time
> like VM time, shred time, real time, tea time and what not... :)
> There should only be one time. VM time. I think of it like a kind of a big
> bang time... everything else is relative. (Did someone say Einstein?)



Yes, there is only one timeline, no issues there. We have "time" which is a
data type that refers to moments and there is "duration" which refers to
lengths of time. This is fine as well.

What isn't so fine, IMHO, is that durations and moments are linked in ways
that aren't as coherent everywhere as (I feel) they might be. After all; the
basic unit of both is essentially the same (floating point numbers of
samples).

Nobody is -so far- requesting multiple kinds of time, this is a potential
issue on a syntax level, as far as I'm concerned, not a proposal to totally
overhaul how we deal with time.


The main issue I have is that we can only define objects of type "time" in
terms of "now" and there is no other way to define such objects which can
lead to roundabout constructs, for example if we want to refer to the birth
of a running VM.

This is mainly a theoretical issue, as far as I can see for now, but a
interesting one that I feel deserves a look. I'm not sure if and when this
will actually affect something practical.

Hope that clarifies my thoughts,
Kas.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.princeton.edu/pipermail/chuck-users/attachments/20080107/2e4d07d5/attachment.htm