For example, I need to deal with some unix process management tasks. I'm not crazy enough to do them in straight C, so I'm using python for it's strong C bindings, so I can use system calls in C with python level control flow. However, there's a lot to be desired, because dealing with all this concurrency and inherent raciness of the system in a language like python is a royal pain in the rearside.
Similar things with java and performant programming. Some of my major programs at work don't use the java ecosystem outside of log4j, junit, apache commons lang and apache commons collections. They use java, because java is darned fast without being as brittle as C or as nuts as C++. So java adds a lot of really cheap development security.
Designing yet another language that puts the existing features together with just a tiny difference doesn't add anything. I've enjoyed looking at experimental languages, and I've enjoyed tinkering around with ideas myself some time ago. But by now, I've accepted I'm not smart or talented enough to make something worthwile, and it will be some time until an interesting new breakthrough occurs.
My problem is quite the opposite: when the languages are so different from what already exists, people (and even myself) have trouble thinking about how adoption would even occur, at least in the short term.
The first step to some kind of adoption is that I can get it running with an "aptitude install crazy-foo-language-dk", maybe after adding a repository. After that, crazy-foo or crazy-fooc should be interpreter and/or compiler, and that's that.
But, when I came to write CL for a project that needed gui, web scraping, using APIs etc (i.e. I wanted to connect it to the outside world) I was stumped. Unlike in e.g. python, there was no nice gui library that I could just use. Now maybe people will reply telling me that something exists, but compared to python, the difference in ecosystem is massive. (possibly some commercial CL environment has this, but I am unwilling to pay for something that is closed).
I really want to use CL for more than just toy programs, but I don't/can't.
So, when you are asking yourself "Is it worth it? Whose lives will it change?" - the answer is mine, but only if you think you are able to fix this.
Good luck.
Html Parser: http://www.cliki.net/cl-html-parse
Http client: http://weitz.de/drakma/
I'm not sure why you are/were stumped? I could search for qt bindings for you, but I'm sure they exist and are pretty mature.
I'm not saying it is impossible, rather that it is hard work compared to e.g. python. I recall looking through a few promising google results and finding e.g. that the library is no longer maintained, has bits missing etc. With python, when I use e.g. beautiful soup, it just works - I lose almost no time getting setup. When doing a project that needs many external libraries, that makes a big difference.
e.g. the first item on http://www.cliki.net/GUI has its last news item from 2007, and the mailing list gives 'not found'. But there is a new(er) github page with the comment "I cleaned up the library just in case I needed GUI in Lisp, but it turned out that I did not. Hence, the primary extent of my testing is running test-gtk:gtk-demo application. Bug reports and/or patches are welcome."
Since the parent was talking about ecosystems etc, I thought my experiences were worth mentioning - "Python novice uses python to get stuff done" but "CL novice gets lost in old/unmaintained libraries, gives up and sticks with c++ & python".
I mean no disrespect to the people who write these libraries etc., I am just describing what I found as a novice in CL.
One can for example also use Clozure CL for advanced GUI stuff, but that is best on the Mac.
For example http://opusmodus.com is a composition software using Clozure CL.
Other composition environments are written with LispWorks: PWGL, OpenMusic, Symbolic Composer, ScoreCloud.
http://repmus.ircam.fr/openmusic/home
I wonder whether the (I'm told) good commercial lisp offerings have hampered lisp adoption, in that people who use lisp a lot, pay for commercial offerings, leaving the free stuff less used and maintained? Then newcomers, who don't want to pay to do hobby projects in a new language, are discouraged by the less maintained free stuff.
This is why ecosystems are, for most people, more valuable than intrinsic language features: Tribalism along the lines of "We have X thousand people backing us up"/"X thousand devs can't go wrong". People don't care about monads or macros, they care about feeling like they're part of a large community.