Hi, I am the upstream maintainer of SCOWL. I have released a new version of the speller dictionaries (and related wordlists) created from SCOWL. New official releases of SCOWL itself will not be created for some time now as the underlying architecture has undergone a complete overhaul and it will be a while before things stabilize. I am writing to request that the Debian source package be updated to use this new dictionary release. I suggest you check out SCOWL directly from Git at https://github.com/en-wl/wordlist. You can either just check out the 'v2' branch to always get the latest version. Or, if you want to track dictionary releases you can use the associated tag created for each release. The latest one is 'rel-2026.02.25'. Your scripts will also likely need to be updated. From the README: SCOWLv2 is a complete overhaul of SCOWL and nearly everything changed. However, there is limited backward compatibility support via the mk-list script. If you used mk-list in SCOWLv1, it will produce the same results. If you created word lists by combining files in the final/ directory your scripts will need to be rewritten. Please use the word-list command of the scowl script to get the word lists you want. If you are using the word-list command please note that the variant levels have changed. The original 0 level is now levels 0-1, the original 1 variant level is now 2-4, level 2 is now 5-6, level 3 is now 7-8, and level 4 is 9. This mapping is also available in the variants_levels table in the database. I will be happy to help with this process. Kevin
On Thu, 26 Feb 2026, Kevin Atkinson wrote:n Thanks for the information. Can you let me know how we'll know that things have stabilized enough for a release? Are you planning on making releases on github from now on? [Looks like the 2026.02.25 is the first release there.] That will help us notice that you've made a new release. [We typically track a specific release, but I can also track your git repo going forward to help with cherry picking and replaying any patches that we have on top of it.] Should we also plan on forwarding any issues to you as github issues (until you tell us differently)? It'll probably be a few weeks before I finalize updating the packaging in Debian for this new release.
On Sat, 28 Feb 2026 08:43:07 -0800 Don Armstrong <don@debian.org> wrote:
Yes, but I do not have an ETA right now.
Yes.
I would track git moving forward through. For stability you can use
the tag associated with the release.
Please.
Okay. I was looking over the source release and I noticed a few
things.
The copyright needs to be updated.
The mk-list command has been ported over and can now output UTF-8,
which can save you some iconv calls. It may be better to use the new
scowl word-list command though.
The scripts in speller/ have been ported over. Do note that they still
mostly work with ISO-8859-1 encoding.
I notice you hacked the `make-hunspell-dict` script to use size 70.
60 is the recommend size for the spell checker dictionary. Please see
the generated Hunspell README for the differences between the default
and large size and why I recommend the 60 size.
If you still want to use the larger size I recommend you just use the
large Hunspell dictionaries created. The larger Hunspell dictionaries
also include common variants in which you will not get by hacking the
size.
A better option may be to package both the default and large
dictionaries and provide some mechanism to select the one to use by
default.
With regard to the /usr/share/dict/ files/packages. Note:
* the default uses size 50, I recommend you bump that to 60.
The 50 size does not get a lot of attention and is not very useful.
* The insane size is gone, the scripts will still work but it will
include a lot less words. I recommend you drop it.
* Some words at the 80 size were bumped to the 85 size; you might
want to change the huge size to 85 to avoid dropping words. Please
see the README for the difference between these sizes.
* you use `-v2` (which translates to up to level 6 (acceptable)).
I do not recommend this as it will include uncommon variants. If
you want variants I recommend you use `-v1` (which translates to up
to level 4 (common)) My personal recommendation is to not include
variants until the large size, but it really depends on what people
use those files for.
Given all the above I recommend you change the SIZE related var. in
the `rules` file to:
SIZES=small "" large huge
SIZE_OPTIONS_small:=-v0 35
SIZE_OPTIONS:=-v0 60
SIZE_OPTIONS_large:=-v1 70
SIZE_OPTIONS_huge:=-v2 85
This will not include most variants until the large size, and then at
the huge size let in the acceptable ones.
If you rather include the variants I recommend this:
SIZES=small "" large huge
SIZE_OPTIONS_small:=-v1 35
SIZE_OPTIONS:=-v1 60
SIZE_OPTIONS_large:=-v2 70
SIZE_OPTIONS_huge:=-v2 85 # or maybe -v3 (level 8, archaic)
# to let in everything
This will let in the common variants up to the default size and then at
the large and huge level let in the acceptable ones.
Please see the README for a detailed explanation of the variant levels.
If you want more fine grain control (for example to also include
`uncommon` variants without also including archaic ones) you will need
to switch to use the word-list command from scowl utility.
Kevin
On Sat, 28 Feb 2026 08:43:07 -0800 Don Armstrong <don@debian.org> wrote: I misread that the first time I answered Sorry. I can't really tell you as I don't have a handle on it myself. Sorry. For now my suggestion is to create the dictionaries and wordlists but not to package SCOWL itself. If anyone wants to use SCOWL it's best they just use it directly from Git. Please do update the dictionaries, however, as there has been a large number of improvements. Also I have been doing through the patches: deprecate_reprized: reprize is gone, so this one can be deleted remove_mys: mys is now at size 85 move_sangs_to_insane: https://github.com/en-wl/wordlist/issues/493 fix_hunspell_affix: one part is already done, the other https://github.com/en-wl/wordlist/issues/494 Kevin