Re: [rox-devel] Re: New ROX App - Find (in files)

[Ken H. <ke...@ha...>]

On Mon, March 6, 2006 10:17 am, Thomas Leonard said:
> On Sun, 05 Mar 2006 10:53:55 +0000, Ken Hayber wrote:
>
>> P.S.  - again, it bothers me that you are preferring 0install to
>> manually
>> installed apps and libs.  Using the svn version of ROX-Lib2, I had to
>> modify filer.py to reverse the order of 'try 0launch' vs 'try rox' to
>> stop
>> things from trying to download Filer all the time.  So currently I am
>> running modified versions of System, Session, Filer and ROX-Lib2 just to
>> prevent this behavior.  This is getting un-fun.
>
> Didn't '0launch --feed /path/to/ROX-Filer.xml' sort it out? What went
> wrong? Do you get any error messages?

Not for me. I wrote about this before, but I guess you didn't see it.

My main problem is that I only want to have 0launch available for
supporting my apps.  Just because it is installed doesn't mean I want to
change the way the rest of my system works.  To me, the --feed thing is
just like having to opt-out of something I didn't ask for in the first
place.  (OK, maybe from your perspective I opted-in by installing 0launch
in the first place.  But I'm only doing it because others asked me to
support 0launch for my applications, not because I want to use it for
myself.)


> There are problems both ways. We have these groups of people:
>
> 1. People who don't want and didn't install 0launch. They're fine.

Yup.

> 2. People who want to use and installed 0launch. Also fine.

Only fine if they want to use 0launch as the default for their system.

> 3. People who installed 0launch but didn't want to use it. Have to add the
> feed, then should be fine (unless it didn't work for some reason).

It may work, but I think it is a pain and this is my main complaint.

In EVERY case I have found so far, the behavior I want is a simple matter
of reversing the order for what is tried first (0launch-local becomes
local-0launch).

>
> 4. People who installed 0launch and want to use it, but already have
> an old 'rox' in their PATH. Currently fine, but would have problems if we
> changed.

This is my point.  I already have most things installed manually and I
like it that way.  For ANYONE who already has gone to the trouble to
install something WHY would they want to download it AGAIN?

> Imagine you're a user on a university machine. 'rox' runs ROX-Filer 1.2.0.
> You get ROX-All and run './rox'. ROX-Filer 2.4.1 downloads and runs. You
> click on Edit. Edit downloads and runs. You click on the Up button and
> suddenly you're in ROX-Filer 1.2.0.

Can't this be solved using 0alias?  Won't that let this individual user
replace the /usr/bin/rox with their own copy (using 0launch)?

It becomes opt-in that way, doesn't it?

> Some apps won't work at all (e.g., Wallpaper).
>
> The possible problems are:
>
> Current system risk: Users download an extra copy of ROX-Filer
> unnecessarily, but everything still works fine. Naive user wastes a
> minute of their life downloading the copy, expert wastes a minute
> registering the local copy.
>
> Changed system risk: Software doesn't work at all because it runs a
> version that's too old. Naive user can't fix it. Expert user has to waste
> several minutes of their lives downloading a new version and installing
> it manually.
>
> (I know there are a few platforms where we don't have a binary yet, but
> users on these platforms generally know what they're doing, and if people
> really cared about it they could just submit a binary. It's still pretty
> easy to select the local copy.)

Well, I guess my only recourse at this point is either:

a) give in and use the --feed thing
b) stop supporting 0launch for my apps
c) write a script to enable-disable the 0launch system just when I'm
testing support for my apps.
d) keep patching my local copies of things (ROX-Filer, ROX-Session,
ROX-Lib, and counting)