On Nov 8, 2005, at 12:29 PM, Bob Ippolito wrote:
> On Nov 8, 2005, at 11:59 AM, Schollnick, Benjamin wrote:
>>>> It's well worth it for me. I reduced the Auditor
>>> application loading
>>>> time by roughly 26 seconds... That uses wxPython,
>>> PythonCard, and a
>>>> variety of other technologies, so it's no light weight... (And is
>>>> roughly 5-6 Meg compressed in a ZIP)
>>> That's totally bogus, modulefinder has nothing to do with runtime.
>>> If your application starts faster, it's because Windows is caching
>>> something so you spend less time waiting on disk.
>> I can't discount completely Windows caching something that is not
>> suppose to.... But these applications are containing in a
>> self-contained, self-unarchiving ZIP file....
>> So the only change I made between 3.10 & 3.10a was the addition
>> of the
>> ModuleFinder patch...
>> Now, there is defintely a speed increase in v3.10a. Since the only
>> change I can attribute this to is the ModuleFinder patch, I must
>> it is the MF patch...
> Well, it doesn't make sense. modulefinder is used entirely at
> build time, and is only used to determine the dependencies of an
> application. There are three possible explanations:
> 1. it's not getting faster
> 2. something else changed
> 3. it's getting faster because the new modulefinder is broken and
> some dependencies that your app wants but doesn't need
and *the result is missing* some dependencies that your app wants but