Re: Loading a package applies automatically to future sessions?

[Stefan Monnier]

> SM> Hopefully John will chime in to correct me, but my vague understanding is
> SM> that his 10-20s were not due to package-initialize, but rather to "manual
> SM> installation" methods and configuration.
> It's not entirely due to package-initialize, but that accounted for about 5
> seconds of it.

Hmm... since my 200 packages take 0.9s on my Thinkpad T61, there's
a good chance that those 5s with your <400 packages are linked to some
poorly packaged packages (more specifically, packages which do too much
work in their <pkg>-autoloads.el, most likely leading not only to a slow
startup but also to undesirable side-effects for those users who
have the package installed yet don't want to use it).

Maybe we should make more efforts at documenting/promoting good
practices in this regard (tho, we should likely first try and figure
out where those 5s are spent, to make sure we don't bark up the wrong
tree).

I'd also be interested to know how much time it takes your machine to
load the "package-fastpath" file that my patch builds (both in source
form and in compiled form).  To distinguish the time taken by the
packages's activation itself from the time taken by package-initialize to
find and fetch that package activation code.


        Stefan