> 
> 
> Try the following hack:
> 

[snip]

Is there anybody out there who wishes to comment on the original
problem, not repeat the change-unknown-into-zero-so-that-the-problem-
is-not-visible hack. ?
(I would post to rrd-users if I wouldn't know this ...)

Don't get me wrong, I do appreciate Alans input however it is not
what is necessary for a solution.

Problem:
We have three directions we graph in. Normally when using only AREA
or LINE, you stack values on top of each other in the Z direction.
If one of the variables is unknown this is no problem for the others.
All variables except for the unknown one get graphed.

When using STACK, one does not longer stack values in the Z direction
but into the Y direction.
When doing so, the start of the area is no longer zero but depends on
the values graphed more early. A technical problem occurs: If the previous
value is unknown, the top of the already graphed values changes from a
number into an unknown. This makes all variables starting from the
unknown one disappear from the image.

My problem with this is *not* that I can't work around it. That's easy.
The problem is the way it is behaving.

Either you see STACKing as adding all values and therefore it should
result in an unknown at that time (#1+#2+#3 is unknown -> graph nothing)
or you look at it as I did in the beginning, without any relationship
and therefore all known values should be graphed. The current behaviour
will display #1 and won't display #3, so you get different results at
different ordering of values.

The purpose of this discussion is to define the correct behaviour of
this. In short, I think the current behaviour is wrong and see two
different ways that it could be defined. I discussed this with Tobi
and suggested to ask for preferences.

My preference would be to skip the unknown entirely (as with normal
area and/or lines) so that #3 gets stacked on top of #1.

Now *you* have to provide input about your preferences (preferably
with the 'why' part) so that version 1.1.x can even be better than
version 1.0.x ...

Regards,
Alex

--
* To unsubscribe from the rrd-developers mailing list, send a message with the
  subject: unsubscribe to [EMAIL PROTECTED]

Reply via email to