[Eric] Request: easier install process, please

[Arlo Belshee]

I really don't mean to whine. I'm actually only sending you this email in
the hope that something will get fixed.

Also, this email is long. Sorry. However, this is the only time that you'll
hear about my complaint - because I run into problems long before I become
invested in your application. Thus, I want to give enough background that
you understand why I, and probably a lot of people like me, don't use Eric.


Background


I first decided to try Eric in 2003 or 2004. At the time, I was choosing my
Python IDE. I have since re-evaluated IDEs several times; Eric has always
been on my list. However, Eric has never won - or really even been
considered on its merits - because of one critical, overriding, and blocking
flaw.

Installing the damn thing sucks.

To me, and IDE is an application. It has competitors. Thus, I try it, and
several competitors out to see how they work for me. At this point, I'm not
invested much in any one application - as is typical for a potential user.
Thus, as is also typical, I want to quickly get in and see what the app will
do for me. In particular, any installation process only gets in my way - the
only thing it can cause me to do is to give up on that application and only
evaluate the competitors.

That's exactly what Eric's install process has done to me, 4 times now.
Honestly, I only keep trying it because Eric claims to have rope
integration. That's my only point of dissatisfaction with my current tools.
In other words, you've already lost, but this one feature makes me willing
to offer you a shot at best 5 out of 9.

Basically, I've got the following set of invisible requirements:

1.       The app should be an app.

a.       It should not mess with (pollute) any system resources (such as my
python install). Any dependency you install centrally means that I can now
miss that dependency when one of my applications has it. Thus, your IDE
gives me the chance to introduce a distribution bug in my application - and
those are one of the hardest types to detect and most embarrassing to ship.

b.       It should also not depend on system resources. I am not going to
worry about your app when I update my system. If Eric breaks, I'll just stop
using it - there are always other options.

2.       As a human, I tell my computer my intent. It then does it. I am not
an automated script.

a.       Thus, your installer should instruct my computer how to do the
install. I should only need to say "yup, install Eric," and then everything
should work. You really, really, really don't want to force me to learn
about your internals before I can even install your app. At this point in
the process, knowing that Eric is written in Python is a negative. I don't
have to know what Firefox is written in to use it. I only need to know that
when I want to extend it - and even then, not really - I just need to know
what language its interpreter runs.

Also, there are a couple of things I really don't care about. Use these to
meet my requirements.

1.       I've got a fat pipe and lots of disk space.

a.       I don't care about download size.

b.       I really don't care about install size. I don't care about install
redundancy. Install a whole new copy of Python if you want. Heck, install 2.
Install one of every minor version from 2.2 through 3.1. As long as they're
inside your app's directory & don't pollute my system, I could care less
what you do.

Remember: this is an IDE. It's going to a non-server machine. I've probably
got about half a TB lying around even after installing my games and music
files. Even on a laptop, I've probably got a few hundred GB free these days.

So, with the above knowledge of what I'm looking for, here's the experience
I have every time I try out a new round of IDEs:


PyDev


Ugh. Another Eclipse-based IDE. OK, fine. Maybe this time it'll be worth it.

1.       Get Eclipse.

2.       Hunt through eclipse for the right way to get a plugin (can never
remember as I only do this once a year or so).

3.       Look around online. Find answer.

4.       Install PyDev extension plugin.

5.       Start using. Discover that most of the advertised good stuff isn't
actually there. It's in the PyDev extensions.

6.       Look up PyDev extensions. No free trial. More install pain.

7.       Well, perhaps I'll try those out later. Let's try a competitor
first.


Komodo


1.       Sign up. Hate to make an account just to get a trial. Oh well.

2.       Download.

3.       Run Installer. Wait, and app opens, ready to explore.

4.       Explore app. This is pretty nifty. Certainly a strong contender.
Let's see what other options are out there. In particular, I'd like
refactoring. Hey, Eric claims good integration with Rope. Let's try that.


Eric


1.       Hit the download page. I have to what! You've got to be kidding me!
OK, fine, I'll try it.

2.       Download the app. Download 8 dependencies (I want to see the
"optional" features, after all - they're the only differentiation between
Eric and the competitors).

3.       It corrupts my system python? Screw that. That'll give me bugs in
my apps.

4.       Well, I could get around that with a virtualenv or non-system
Python install. Yeah, I'll try that if the other competitors all suck.

5.       Bail.


Wing IDE


1.       Download the installer.

2.       Run it. Wait. App loads, and asks me whether I want to start a free
trial. Click yes.

3.       I'm in and exploring the app - without any pain already built up.
Means I'm much more likely to have good impressions about their features.


Evaluation


OK, so I've got two of them that look good. Should I go back and try one of
the others? Well, both claim to add refactoring support. I'd really like
that. But both are a pain to install.

I don't want to waste time right now. Let's just use these two and get back
to work on the thing I actually care about - my project. Maybe I'll look at
the others in a couple of weeks.

A couple of weeks pass.

I don't even think to go back and try the others. I've learned how to make
one (or both) of Wing and Komodo do what I want. It never even occurs to me
any more to look for an IDE. I've solved that problem. Eventually, I see the
downloaded files & PyDev's Eclipse instance lying around. I laugh and hit
the delete key.

In summary, I'd really like to try Eric. Rope integration makes it worth
writing you this email. However, it doesn't make it worth corrupting my
system Python, going through a long, manual install process, or trying to
adapt that install process into one that doesn't corrupt my Python.

I hope, one day, to be able to evaluate Eric on its own merits, rather than
just on the abject failure of its install process.

In the mean time, I've got 2 IDEs that are good enough, so no real skin off
my nose.

Arlo

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.riverbankcomputing.com/pipermail/eric/attachments/20090723/538ebfbd/attachment-0001.html