Btw, I don't think a Gemfile adds anything for a bundled production
app besides that our shim would have to support the DSL. I think the
dev should just use ‘require "bananas"’ in her code and it should just
work. I.e. the Gemfile is I think only needed in development and at
release time.

On Sat, Jan 29, 2011 at 11:08 AM, Eloy Duran <eloy.de.en...@gmail.com> wrote:
> Yes, what we need is a tool that resolves dependencies at release time
> and bundle these in the app. The app also have a shim that makes sure
> that if any of the bundled dependencies call Kernel#gem it doesn't
> break. I.e. on the clients machine rubygems should not actually be
> loaded to ensure as fast as possible startup times.
>
> When I recently spoke to Yehuda he said that bundler now has a
> `standalone' feature which provides exactly the aforementioned
> requirements. I haven't tried it, nor do I suggest we use it, but
> having written my own dependency resolver in the past, it might be a
> nice idea to at least try it.
>
> @Laurent: Do you know if Rails 3 will be shipped with future OS X
> updates? If so, bundler probably will come with the system too, so
> it's not really a problem to depend on it. Especially since you don't
> *have* to use it if you wish to do it another way.
>
> To the rest, I would love to create a patch for this but don't have
> enough free time atm. So instead of discussing this again and in the
> end not having anything, I propose somebody that currently needs it
> and has the time to create a patch that does the described
> requirements. Only then will we be able to really judge whether or not
> the full bundler lib is what we need or if we can write our own (or
> extract) subset of it and continue the discussion from there.
>
> On Sat, Jan 29, 2011 at 5:24 AM, Caio Chassot <li...@caiochassot.com> wrote:
>> On 2011-01-29, at 02:11 , Matt Aimonetti wrote:
>>>
>>> IMHO adding Bundler as a dependency would be a huge mistake. First, Bundler 
>>> code is overly complicated and tries to do way too much.
>>
>> (You forgot to mention slow.)
>>
>> I think I agree here. I just like the Gemfile format. Don't think we need to 
>> reinvent the wheel.
>>
>>
>>> Trying to resolve the dependency tree is not something you should do in 
>>> prod or on a client machine. This belongs to the dev environment.
>>
>> Maybe your argument backfires here, because resolving the dependency tree at 
>> compile time to decide which gems to bundle in the final app is actually 
>> quite useful. In fact, it's what you suggest right after:
>>
>>
>>> The way I see it, we just need macdeploy to unpack the gems you need & 
>>> their dependencies as well as modify the loadpath. By doing that our prod 
>>> apps don't even have the need to load or rely on rubygems.
>>
>> Yes. In fact not loading rubygems on bundled/compiled apps is probably 
>> something to strive for. I can see things getting messy with unforeseen 
>> conflicts between app gems and user gems.
>>
>> _______________________________________________
>> MacRuby-devel mailing list
>> MacRuby-devel@lists.macosforge.org
>> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
>>
>
_______________________________________________
MacRuby-devel mailing list
MacRuby-devel@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel

Reply via email to