Re: [Libphidget-devel] serial numbers

[Jack S. <js...@ja...>]

On Thu, 2002-09-12 at 02:33, Vadim Tkachenko wrote:
> According to fit...@ph...:
> 
> > > Then, the string serves as good as the number, though you can a) add more
> > > significant information to it (like, the release variations, or specific
> > > options included into an instance of a phidget), b) extend the address space
> > > (though this is rather theoretical at this moment).
> > >
> > > How much of the existing code will break if the serial happens to be a
> > > malformed number?
> > 
> > Well, it doesn't bother me very much, because I just release a new Windows
> > version and tell everyone to go use it.  It's nice owning your own
> > sandbox.  I just try not to change it all the time..  As long as my
> > library is smart enough to figure out all the old phidgets, I don't worry
> > about broken code.
> > 
> > I don't think it's really worth mucking with, because USB provides other
> > mechanisms for device versioning, and so forth.  Particularly with the HID
> > devices.
> > 
> > With the soft phidgets, all that data, including the serial number,
> > version, .... extraneous information like programmable logic
> > device firmware versions, get stored into a fat EEPROM.  And it makes no
> > sense to cram everything into the serial number when it can stay
> > structured?
> 
> True. My rationale, though, was: Occam's Razor. Why should I parse the
> serial to number when it is *enough* to treat it as a string?

treating it as a string might be fine, but if Chester will give us a
heads up before he starts assigning alpha numeric serial numbers we can
change the code before he starts selling phidgets with such.

> 
> One reason I might think of: memory. But is this reason good enough? Not to
> me - how many phidgets will I have on the adapter, and how many bytes will I
> lose if I store serials as numbers? Definitely less than amount of code
> necessary to parse it...

Memory isn't an issue, you can only have 128 USB device (I think), so
128*256 (Max string length), what's that 32K, think I could spare that
worse case.

> 
> Performance - how often do I have to compare the serial with something?
> Don't think more than once - at initialization.

Actually, since this is the unique ID for all devices it is used QUITE a
lot, but I don't think a string compare will be too expensive.

> 
> Just thinking aloud...
>