Re: How to make GNU Guile more successful

[Erik Edrosa]

On 03/04/2017 06:41 PM, Alejandro Sanchez wrote:
> If I may add my two cents as a Scheme layman: the question is not so much 
> about making Guile more popular, but about making Scheme itself more popular.
> 
> One big reason for Python’s popularity is something I haven’t seen mentioned 
> in this thread so far: if you know pseudocode you basically know Python 
> already. Of course this is hyperbolic, there are a lot of finer details to 
> Python, but the superficial simplicity of Python makes you think you already 
> know the language and that you can get started right away. By the time you 
> encounter the more arcane aspects of Python you have already invested enough 
> time into it that you will put up with learning something new.
> 
> Scheme on the other hand is weird; weird in a good way, but still weird. For 
> me the reason for picking up Scheme was working my way through SICP, but even 
> before that I had heard of the Lisp family. Every time I tried to find out 
> what the fuss was about Lisp all I could find was nebulous concepts like 
> “it’s a programmable programming language” without explaining what that even 
> meant and why I would ever want to use it.
> 
> Of course once I got over the weird parentheses and I understood why they are 
> actually a beneficial aspect I liked it. Yet, even after getting into Scheme 
> there are problems. The r5rs standard is pretty limited, and so every 
> implementation does its own thing. There was not even a standard module 
> system until r6rs. Unfortunately r6rs never got fully adopted by any of the 
> major implementations, so we are back to where we were with r5rs more or 
> less. r7rs-small looks like it could finally fill in the most glaring gaps, 
> and r7rs-large is still nowhere done.
> 
> All this is very frustrating, and getting into a new language is a long-term 
> commitment to learn something new. When faced with Python, Ruby, JavaScript, 
> Lua and Scheme, it’s easy to go with the flow and just use what everyone else 
> is using.
> 
> I think someone here mentioned the lack of a proper Scheme IDE, other than 
> Dr. Racket for Racket. I don’t use IDEs anymore, but I can see how that can 
> be a problem for other people who only want to do intermediate scripting 
> rather than write entire applications. Writing a full IDE would be quite a 
> lot of work, so perhaps integrating with existing IDEs would be better. Think 
> something like Geiser for programs other than Emacs. There is the “language 
> server protocol" by Microsoft, the idea is to have a standardised protocol 
> that can be supported by a language server and an editor application.
> https://github.com/Microsoft/language-server-protocol
> 
> I know it’s Microsoft, but this is just a specification. With this protocol 
> one would write a Scheme server that receives instructions form the editor 
> via RPC and sends the result back to the server via RPC as well. This would 
> allow using the same Scheme server regardless of the editor (Emacs, Vim, 
> Neovim, Eclipse, …). I had wanted to write something like this for Neovim 
> myself, but I don’t know enough of Scheme yet.
> 
> I hope these ramblings of mine offer something useful from an outsider 
> perspective to you guys.
> 

I think one issue when you want to compare languages like Python to
Scheme is that these languages are dominated by a single implementation.
Python, Ruby, Go, Rust, Clojure, and many others have a single dominate
implementation. The compatibility of libraries isn't something devs
usually have to worry about (except for breaking changes in the language
like Python 2 to Python 3). I agree with how r5rs standard is pretty
limited, languages with multiple implementations tend to have pretty
large standards so that these implementations are somewhat compatible. I
am liking the direction of r7rs and I hope it will bring scheme
implementations closer together. Probably the other option for scheme
would be for everyone to get behind a single implementation like Chez
Scheme.

Of course that isn't the only thing needed to make Scheme or Guile more
popular. I believe a programming language needs a somewhat unique and
powerful library or framework, an awesome application to show off the
power of the language, good tooling, and some luck.

The language server protocol is pretty cool and I would like to see a
lisp implement it. An IDE for a lisp obviously needs to implement more
features to be useful, I believe being able to easily interact with a
REPL is one of those.