Is Tcl still relevant with the momentum that Lua has gained?
On the other hand, there are enough applications in enterprise and manufacturing where Tcl/Tk applications run happily along and there's no good reason to replace them. Your nuclear power plant might have a neat little charts GUI done with Tk's canvas and Tcl scripts to tie all kinds of sensors and gauges together. Your automotive supplier might still do CAD/CAM file conversion/distribution/deployment with a Tcl server.
For new applications, it's probably a hard sell these days. Don't get me wrong, I personally think it's a great little system, especially when you're in an 'ivory tower' situation where you control the whole setup and don't need a lot of interop middleware and libraries. It's easy to create simple GUIs for all kinds of tasks and I've rarely seen systems that make it easier to leverage 'non-programmers', i.e. get some user scripting done.
For ugly, dirty hack jobs like web programming, I'd rather go with something more standard or -- if I feel inclined to reinvent the wheel -- with something more powerful and/or rigid (whether that's CL, Haskell or Ada...).
Although weirdly enough, web client programming is certainly something that checks both the 'ivory tower' and 'non-programmer' boxes...
Tk seems pretty pointless now. Wx obviates it, in my opinion.
Dunno about Wx. Last time I looked at it, it was a horrible morass of classes and constants that actually made me long for Swing...
So the momentum is about the same and therefore both are relevant.
No doubt it's not going to be dethroned in any way from the numerous industry tools (most CAD systems support it for example) but if you want to embed a scripting language, is there any reason to use Tcl over Lua these days?
http://www.nist.gov/el/msid/expect.cfm
Exploring Expect by Don Libes, the principal author of Expect, is a classic among tech books:
As an example, a few years back some coworkers and I were interviewing a candidate and he used a lookup take to avoid having to constantly calculate sin and cos. We wrote some tests to discover what the performance improvement was and found out there was none. It was slower to read a value out of an array than it was to calculate the sin or cos of a value.
Tcl might have been a good option back in the day, but they just don't keep up with everyone else, both in terms of performance and feature set.
Have you even looked at it recently? Just curious.