I have a soft corner for TCL, because of its idiosyncracies and weird dynamic stringy playfulness. I would be wary of using it professionally but to play around for the fun of it ... it's just too much fun. Upvar and uplevels are crazy.

Python is that kind of cool one behaves when visiting parents of your would be spouse for the first time. Tcl on the other hand is like playing one's secret exclusive and somewhat dangerous games with kid school bestie.

Tcl/Tk pioneered the notion of a scripting language as a library. It got threading right. You can run multiple independent Tcl interpreters in the address space of your process and they can exchange messages. Entirety of interpreter state is encapsulated inside the interpreter object, no globals. To a user an it is just a pointer to an object.

No need for GIL, no need for serialization/deserialization (pickle/unpickle) ... just within process memory copy (or no copies in case data is immutable).

> I would be wary of using it professionally

This is funny because so much Silicon CAD automation is written in TCL. I mean the most advanced technology in the world depends on TCL in its pipeline (also Excel VBasic!) (let's not talk about COBOL, that's in banking)

TCL is the original extension language, and in the very hidebound electronics CAD world, has a very long tail.

Not to sidetrack that much, COBOL is now in ISO 2023, even has OOP features, there are ways to do microservices in it, and nice modern IDEs as well.

Additionally maybe we should stop complaining about its verbosity, given the amount of English based programming going on nowadays.

I know.

I would hazard a guess that a significant fraction of HNers would have encountered TCL in silicon CAD and took with them a frustrating experience.

> Entirety of interpreter state is encapsulated inside the interpreter object

This was their attempt to position Tcl as an alternative to JS. The sub-interpreters were introduced as controllable sandboxes for running untrusted code with reduced privilege. Instead we got a web scripting language that allows untrusted code to interact with your USB devices. Python sub-interpreters are a joke compared to what Tcl has.

Are you sure ? Tcl is almost a decade older than javascript although subinterpreters came later. Even before subinterpreters you could have multiple Tcl interpreter instances in your code.

One could call Tcl_CreateInterp() multiple times in your code and each call would return an independent interpreter.

Safe interps were designed c. 1993 by Nathaniel Borenstein and Marshall Rose to be able to safely run code received through email[0][1].

[0] http://www.beedub.com/book/2nd/interp.doc.html

[1] https://www.guppylake.com/nsb/pubs/ulpaa-94.pdf

if anyone was curious! "soft spot" has an Indian English "soft corner", so says website dot com

Very interesting. I was not aware at all that this was an Indianism. One that I like quite a bit is the word prepone.

My uncle, who worked for oil companies all his life, always had the opinion that the Indians use english better than the English. :-)

'prepone' is an example of why I sometimes agree. Most indian guys I know also up the baudrate of english so much that they get twice the information flow compared to british english. With the disadvantage that both sides of the conversation need to speak indian english or you'll get protocol errors :-D

Sort of interesting how Python has got rid of the GIL. Serialization of data securities is still something relevant though and I’m not sure if you should use pickle still?

Pickle is insecure by design, but definitely convenient if you can trust the input.

If you need something safe then you really need to identify your requirements and your threat model first anyway; I wouldn't blindly reach for a one-size-fits-all solution in any programming language.

(Although I would do a quick check to see JSON or TOML is sufficient and practical.)

Pickle is likely still handy, as long as you created the data that you're deserializing. Zope's object database was a big pickle file, as I recall. But yeah, not a good idea if you're cracking open user-supplied files in any way.

Ha! People still remember Zope!

That was my first web kind of framework. The naivette of Zope still amuses me endlessly to this very day.

It tried to be cool, made ALL the wrong choices, and died silently with people just moving on.

"wary of using it professionally". what does this even mean? you can use anything professionally if done so in a professional manner. it's just a tool, use it correctly and within limits.

You can find yourself having to work with quite unprofessionally written code. Not an uncommon experience.

Tcl is certainly a bit weird and takes getting used to, but I have found it worth learning its idiosyncrasies, particularly because my applications use sqlite, which fits very naturally with Tcl.

From D. Richard Hipp [1]:

"SQLite is a TCL extension that has escaped into the wild.

"The design of SQLite was inspired by the design of TCL, both in the way it handles datatypes and in the formatting of its source code. The index use case for SQLite was in a Tcl/Tk application for an industrial company. From its inception, SQLite has always depended heavily on TCL. These days, SQLite no longer uses TCL internally and can be run separately from any TCL interpreter, and yet the SQLite development process still depends heavily on TCL."

[1] https://www.tcl-lang.org/community/tcl2017/assets/talk93/Pap...

The biggest early selling point of Tcl was Tk, as a relatively easy way to make good-enough GUI programs with open source on Unixen and the X Window System. Even much easier than the easier-to-use X toolkits like XView with C, and the alternatives only got more difficult from there. There were no Web frontends.

Tcl itself was quite clever as what I'd actually call a string-based scripting language (contrast with Bourne/C-shell/etc. that preceded it on Unix).

Originally, Tcl was among the handful of off-the-shelf extension languages. You write your big application in C or C++, and then you embed an interpreter for a higher-level extension language, for users or the original developer to add on functionality. The options at the time were usually a small Lisp/Scheme, Tcl, Python, or something entirely bespoke and probably quirky and half-butted.

During early dotcoms, there was considerable interest in Tcl for backend work, and it also wouldn't have been the worst choice to put into the browser. Like Python, Tcl would've been more accessible and democratizing-the-Web than JavaScript, and there were already solid off-the-shelf designs and implementations. (Scheme would've appeared slightly more intimidating than Python or Tcl, and was more powerful, which I think was why Tim Berners-Lee preferred Python over Scheme as the people's programming language for various Web purpose. The semantics of the JavaScript we got was more a hurried toy Scheme, plus the simplest object model, and a more intimidating syntax that came from systems programming.)

Here is a discussion from last month:

  Develop Cross-Platform CLI and GUI Tools with Tcl/Tk (cgicoffee.com) 
  112 points by aryonoco 29 days ago | hide | past | favorite | 62 comments

https://news.ycombinator.com/item?id=49515662

Yep, the age of Python and the C-steeped culture of its original developers probably has a lot to do with the choice of Tcl/Tk as a basis for the standard library GUI framework. But even then, there's significant wrapping; and Python's type system is, if anything, even further away from the free-wheeling "stringly typed" Tcl than JavaScript is.

I do sometimes yearn for the alternate universe where web browsers ended up doing Python natively (and PyPy style optimizations became the default, and the dynamism was reined in to an extent where it could offer JavaScript-tier performance).

Aside from the emphasis on "continuation passing style" though I don't get how you're relating JavaScript to Scheme.

So I also love Tcl, but I think it would have not been great as a web language. I think the small amount of type information we can pass via json is just about right, and tcl would have made it impossible.

You cannot encode arbitrary tcl data to a typed object "correctly" because every tcl dictionary is a valid list, and every list is a valid string (numbers are also strings, and so is code). So in a world of "tcl code on the web", both the sending and parsing ends of a 'tcl data stream' would need to know which type they're "supposed" to treat the data as.

> The biggest early selling point of Tcl was Tk

You're not that old, are you? :) Before www, expect was the way to go.

<https://en.wikipedia.org/wiki/Expect>

Expect was amazing - game changing. The things you could do with it couldn't be replicated elsewhere (well, clearly "could", but not reasonably so)

My pet project in those days was writing a C++ X-Windows gui toolkit, and I took a lot of inspiration from Tk. Both design, and "how do I do this" from the source code were invaluable.

Didn't Cisco use Tcl in one of their IOSes? Found it: https://www.cisco.com/c/en/us/td/docs/ios-xml/ios/ios_tcl/co...

Yes the legacy Cisco IOS had a Tcl interpreter built in, you can drop in to the repl with the tclsh command. The command is still there on newer Linux based IOS versions

Y'all remember TkDesk at all? https://tkdesk.sourceforge.net/

I loved it back in the day, tons of fun to configure with code, somehow friendly enough for me to play with in high school with no background knowledge and only some man pages to help me. Coupled with a lightweight WM like CTWM https://www.ctwm.org/index.html it lead to a really performant desktop on the single-core ~400mhz machines we had in our computer lab....

Iirc it was successfully used so in the CGi world (nuke and maybe maya mel)

I think some of the "LAMP" type bundles use Tcl/Tk for their UI, kinda neat.

Tcl/Tk has got to be the easiest GUI system out there; you can get simple stuff going with no effort - nothing I've seen comes close to that simplicity.

Good to see it's getting some modern support.

It hits a sweet spot for a lot of basic UI needs.

Is it still true if I want to custom widget with animations?

I've done animations at the Tcl layer before with composition of Tk objects. for little visualizations its just fine, certainly not AAA

What about oldschool Visual Basic?

Besides the sibling comment, VB.NET is still around, even if not with the attention it once had.

VB + C# with Windows Forms is not much different from old school VB.

And for the pixel positioning complaints, it is about time people learn about FlowLayoutPanel and TableLayoutPanel, they exist since .NET 2.0 (2005).

It was nice/workable for its time, and one wishes that there was some modern alternative, but "RAD" seems to have dropped off the radar.

Some things I've tried/looked into (beyond TCL/TK):

- Gambas --- Linux only, not sure if there's a nice visual UI builder

- LiveCode --- still salty about the license rug pull, wish the openxtalk folks could get some traction

- MoonBasic --- looks promising, but currently trying....

- flet.dev --- the AI-integration in this makes it quite compelling, though I wish that there was an integrated UI drawing tool, https://github.com/raffieeey/Flet-Visual-Builder is self-described as buggy and hasn't seen an update in 7 months

- Lazarus --- still bummed I never got anywhere with Delphi --- if flet.dev doesn't work out, may try it

Tcl is way too smart of a language for its own good. I love how everything is a string or can be hacked on a string, so you can basically do levels of metaprogramming that reach unheard of levels

Not to mention uplevels and upvar.

All of which I used many years to go to write "Snit", a quite nice object system for coding up Tcl objects and Tk widgets. Snit's a TCL library that takes a nice, easily readable description of your desired object type (the instance variables, the methods, and so forth), translates it immediately into standard TCL, and then evaluates that to define your type. Basically, it's a macro system for defining object types. (I say "object types" rather than "classes" because there is no inheritance.)

I loved your 'expand' utility - and wrote a truly heinous Tcl to PHP framework/translation layer using it, because I preferred Tcl to PHP at the time - still do, but the early hatred for PHP has muted - that powered an application at the college I attended for an embarrassingly long time.

Thanks!

Would this be the best place to learn about it

https://wiki.tcl-lang.org/page/Snit%27s+Not+Incr+Tcl

Not really. Snit’s included in Tcllib; find the Tcllib docs and look at the Snit man page and Snit FAQ.

I've always thought of Tcl as a type of Lisp. Clearly, the relationship isn't direct, but it allows a lot of the same things that Lisps do. Plus, the way one nests commands is similar.

And it shares a lot of the same metaprogramming that Lisps have - more so than the vast majority of other languages.

Congrats to the community and the core team. Their dedication to maintaining Tcl/Tk as a gold standard of stability, pragmatism, and lightweight cross-platform development is inspiring.

I once had an O'Really book named 'Perl/Tk'. I mean, I still have it somewhere... I've the feeling reading through it will maybe become my AI recovery therapy at some point?

Perl/Tk is a lot of fun. I wrote a solitaire game called tktk — Tk Time Killer — that appeared in The Perl Journal.

https://blog.gbacon.com/publications/2000-perl-journal-tktk/

In a way its distant nominative descendent, TikTok, is also a solitaire game.

This is very cool! The name alone. :-)

Thanks! It was a fun project.

Perl/Tk is a hard fork and hasn't fully integrated the improvements in Tk over the last couple decades.

That's a bummer. Now I have to start over. </s>

For the full release announcements see:

https://newsgrouper.org/%3C119gpah$3eu5q$1@dont-email.me%3E (Tcl) and

https://newsgrouper.org/%3C119gpbh$3eu5q$2@dont-email.me%3E (Tk).

Many VLSI CAD tools come with a tcl console window. Im gratified to see the continued language support

Kudo’s to Mr Osterhout’s long lived legacy

> Kudo’s to Mr Osterhout’s long lived legacy

Tcl and Tk are fantastic, and if this is where it ended, John Ousterhout would have secured his place in history - this is only 1 piece of his contributions, though - see too:

  * Log structured file system[0]

  * parallel Make[1] (Adam de Boor from Sprite project, not JO hisself)

  * RAFT consensus protocol[2]

  * Various notable teachings [3][4][5]

[0] https://en.wikipedia.org/wiki/Log-structured_file_system

[1] https://man.freebsd.org/cgi/man.cgi?query=bmake&sektion=1

[2] https://en.wikipedia.org/wiki/Raft_(algorithm)

[3] https://en.wikipedia.org/wiki/Ousterhout's_dichotomy

[4] https://web.stanford.edu/~ouster/cgi-bin/papers/threads.pdf

[5] https://web.stanford.edu/~ouster/cgi-bin/aposd.php

I'm sad he's retired.

Here's another plug for his book:

A Philosophy of Software Design/2e

https://www.amazon.com/dp/173210221X

It's unfortunate that Sprite itself didn't really get absorbed into much of anything, and so doesn't have much of a footprint and isn't really remembered outside of LFS and a few even more niche bits.

But I think the Sprite kernel and the Sprite-specific libraries (Pfs, Pdev, Fs_Select, Td_), as well as the Sprite-specific userspace, were pretty incredibly pieces of software given the limited resources they were created with.

For that matter, John Ousterhout's text editor and terminal emulator, mx and tx, were the substratum that Tcl was built on. There's a single library on Sprite, libmx, which mx and tx are based on. Tcl basically grew as the command language of libmx, originally two separate entities, then by mx 2.4 it was fully incorporated:

  eery@cherimoya [1] > cd /sprite/src/lib/mx
  eery@cherimoya [4] > grep Tcl *.c
  mxCmdAM.c:    Tcl_Interp *interp;
  mxCmdAM.c: Tcl_Return(interp, Cmd_BindingGet(mxwPtr->cmdTable, argv[1]),
  mxCmdAM.c: interp->result = Tcl_Merge(bindingCount, bindings);
  mxCmdAM.c:    Tcl_Interp *interp;
  [...]
  eery@cherimoya [10] > grep LIBS ../../cmds/mx/local.mk
  LIBS  = -lc -lmx -lsx -lcmd -ltcl -lX11

Of course, even by 1992, there were multiple versions of Tcl coexisting... :-)

  eery@cherimoya [11] > ls /sprite/lib/sun4.md | grep tcl
  libtcl.a
  libtcl5.a
  libtcl6.2.a
  libtcl6.3.a

I see this praise of tcl/tk a lot on HN, but everyone I know (myself included) absolutely hate working with it in VLSI CAD tools.

If you write Tcl/Tk from scratch for an application and architect it well, it's great. The way it's injected into VLSI tooling is an abomination. I've done both.

This is partially because when Tcl was first incorporated in VLSI tools it didn't have a good way to organize and modularize code. But this is also because the big VLSI tool shops realize they basically have a monopoly on their integration surface and have no incentive to improve it.

It's easier to freeze the integration surface and maintain backward compatibility than it is to move the surface forward.

One can create amazing ergonomic scripting APIs in TCL; I've done that. One can also rapidly create horrors without end. I've seen that, too. With great power comes great responsiblity.

Indeed. If AI did nothing for me but deal with Xilinx's shit, I'd still nominate everybody from Hinton to Amodei for Nobels.

I like it better than the alternatives. The challenges come from vendors who had atrocious in house scripting languages (Synopsys) and then shoehorned their architecture into Tcl. You have to deal with a lot of opaque handles for things that should be native objects.

New version of AOLServer incoming?

Phil Greenspun is still active on X: https://x.com/PhilipGreenspun

Some throwbacks of his from 1999:

https://philip.greenspun.com/wtr/aolserver/introduction-1.ht...

https://philip.greenspun.com/wtr/aolserver/introduction-2.ht...

As a teen I attended a weeklong ArsDigita workshop in the vault of their Pasadena office. It was in an old bank building on Green St. Aeron chairs, Sun Ray thin clients (iirc), a huge circular bank vault door...

I remember being introduced to regex one day, and later the instructor (a Caltech PhD?) being surprised my code passed all his tests in a class review.

Really, a wonderful experience.

I thank my cousin, who lived by the 7Cs, for hosting me and driving so far to pick me up the first night (after I hung out waay too long with new acquaintances in the AD office).

And I remember, after a day of learning Tcl, while waiting for an express bus back to Claremont, listening to a woman belt out opera on an empty Colorado Blvd in front of Vroman's, like a private opera rehearsal.

Such a classy city.

https://github.com/naviserver-project/naviserver has you covered

Now if we could just bring back Navi/AOLPress....

Ah, the memories.

The startup I joined in 1999 had its own application server loosely based on AOLServer architecture.

We had Apache + mod_tcl, IIS with our own ISAPI extension, then in box connections for all major RDMS across all key UNIXes and Windows NT/2000.

Some left to join companies doing Vignette projects, others eventually created OutSystems redoing the same ideas, but with the newly released .NET.

However it was also a lesson in performance issues, and the constant pressure to rewrite Tcl code into C extensions.

I used it when I worked at AOL :)

Was great for rapid prototyping and a/b testing in volume. Not sure I'd pick it ever again for anything, but it was interesting for sure.

When Rails came out, I was like what is the buzz all about, we have been doing this already for several years.

Same here. I enjoyed that time/experience a lot.

I remember using tcl inside pages running under Vignette Story Server circa 2000 before we started with jsp

And OpenACS!

Or Kenny Tilton's celtk or cello?

Anyone have fun stories of things they built with tcl/tk? I feel like there's still room for tiny apps with it in theory, but I tend to have the problem of wanting to stuff "arbitrary rich text" into my things (-> web tech suddenly works quite well)

The cool thing is that even if you take something arguably super modern, like the latest Claude models, it is able to write Tcl/tk applications for you that you might not have done at that level in the past. With tests in tcltest too.

If you understand how generative models and agentic coding works, it's not a surprise. But for some reason it's still mind boggling to think that you can take very old stuff and build impressive programs.

Indeed!

For the simple prompt to Claude Fable:

    let's write a small tcl/tk app for fun.
    a small tiny UI to manage a to do list.

I got this in no time: https://gist.github.com/mathusiast/a57bf0dfa42d8c7008f8882da...

I haven't tested this recently, but I bet it’s equally capable of writing custom Tcl extensions using the C API. Given how stable and clean the Tcl C interface has always been, it’s probably a perfect use case for LLM generation.

Yeah should work. So that makes it possible to do literally anything. Tcl always had one foot in the old days and one foot in the modern era. It's not like ALGOL or anything.

> It's not like ALGOL or anything.

It's a Lisp with syntax sugar and no macros.

Lisp a la Unix way.

The little language that could.

“Tcl tends to get ported to weird places like routers.” (Larry Wall, October 1997)

Tkinter[0] was apparently released in 1994, so mildly interesting he didn't say "Tcl tends to get ported to weird places like other programming languages."

[0] https://grokipedia.com/page/Tkinter

It's not really ported. tkinter just builds a full Tcl interpreter as a Python extension and passes command strings over to it. This is the use case Tcl was designed for. You can get access to the underlying Tcl instance and issue non-Tk commands.

Technically correct (the best kind of "correct").

My next question to myself was what hardware was running Tcl in c.1997 Cisco equipment? What porting effort was required? I understand there was some proprietary feature work for Cisco IOS, but I don't know what architectural work was required.

They don't make them like they used to

I did some Tcl work a quarter century agi, but really if I were in the market for an extension glue language today, I'd reach for Lua first.

I like the concept of Lua, but I ran into enough issues with it's "everything is a table" design that I stopped using it.

How come? I don't have enough experience with either to have a preference, but I'd love to hear the opinion of someone more familiar.

Not the OP but Tcl comes with batteries included and more "full fledged". Lua is just so much more lightweight and so easy to just drop it in any codebase.

The biggest reason to use Tcl even to this day is to write Tk GUI programs.

> The biggest reason to use Tcl even to this day is to write Tk GUI programs.

Well... Tcl on its own is good as a configuration language, creating DSLs, or as a scripting/extension interface, as well as a single common cross platform interface to filesystems, network sockets, and more - so I'd say "the biggest reason" depends on what you're working on...

Interesting! Would that make a different choice for standalone use, then?

I wonder when Linux distros will start shipping 9.0 or 9.1. Even Arch is still stuck on 8.6, despite 9.0 being released two years ago.

9.0 was unfortunately incompatible with a lot of software. Certainly in Fedora we did the work (about a year ago) and have been shipping it for a long time.

Omarchy is a hyped up Arch Linux configuration tool with one specific kind of user in mind (namely its creator). It’s not beginner friendly if your brother hasn’t used Linux before. In fact Arch Linux itself is not beginner friendly. It’s better to stick with Ubuntu.

Many years ago, I worked at a company that used Tcl as the scripting language for its primary product. At one point, a hotshot product manager started agitating to replace it with something more modern. He claimed that Tcl was dying because (wait for it) there were no books being published about it at the time. I guess he was wrong ... :-)

It's in decent company in the Redmonk graphs[1], in the vicinity of Smalltalk, Mathematica, D, and Haxe.

[1]: https://redmonk.com/sogrady/2026/04/14/language-rankings-1-2...

Excellent to read Tk is getting better acessibility support. I love using it for quick little Perl/Tk GUI applications but I'm also slowly going blind from retinal tearing. It's good to see progress in some GUI toolkits when major desktops like KDE have dropped accessibility in their latest releases.

In what way has KDE dropped accessibility in their latest releases? (Genuine question; first time hearing about this, but I do not follow the space closely.)

Wayland.

Last time I checked, there was no Wayland support (other than executing in XWayland). Has it changed?

Wayland support is being worked on, aiming for release in Tk 9.2 next year.

See: https://wiki.tcl-lang.org/page/GSoC+Idea%3A+Tk+Backend+for+t...

glfw! that's a bit surprising.

i'd expect sdl, given wider platform coverage (e.g. mobile), more system integration (e.g. clipboard), and an existing port (https://androwish.org/home/dir?ci=trunk&name=jni%2Fsdl2tk)

glfw is smaller and more focused (SDL adds a TON of stuff, including even its own GPU abstraction API) so the choice was most likely made because of that focus.

I tried to get into tk via ruby-tk.

I ended up really disliking tk.

Using ruby-gtk2 and ruby-gtk3 was much nicer. (Sadly, gtk4 sucks, so that's the end of the story there.)

I kind of want a universal toolkit that works everywhere and is also good. Right now I use ... well, either the web (that's ok, though I hate how I can not easily open local files and so forth, without having to use node), or swing via jruby-swing - which semi-sucks, but it also works on windows and it actually is not that bad. (I could use javafx etc... but it is so much easier to get started with swing). I also use libui-ng which is ok but lacks many features, sadly.

Tcl/Tk has saved me back in the day. The Windows distribution BAWT/Magicsplat are timely updated

..... but they do not come with Next Scripting Framework.

There was a new release on September 16, 2026

In the early 2000s, I used the Tcl/Tk bindings to the VxWorks target management API. The application hotloaded and -unloaded test modules for a hardware in the loop lab. Tcl allowed the first working concept to come to life in a few days. I have no idea how long it would have taken with with C++ library.

I'm familiar with Tcl because many Cisco routers actually support it as a scripting language.

Try https://wiki.tcl-lang.org/page/Showcase .

Also https://cgicoffee.com/blog/2026/04/tcl-tk-develop-cross-plat... (takes a little time to load).

I wonder what the author of the second article would have thought of wxWidgets in their “why” section. It wouldn’t have the drawbacks that they attributed to Qt.

Wow, that sounds really traumatic :(