1 files changed, 49 insertions, 28 deletions

diff --git a/TODO.md b/TODO.md
index dc5a44b..3453ba0 100644
--- a/TODO.md
+++ b/TODO.md
@@ -4,34 +4,51 @@ Make it perphekd!
 
 ## Do these
 
-Add atom feed for single page.  Alas cgit only seems to have ATOM feed on the whole repo, not individual files.
+Convert it to polygLua.
 
-- However, once timestamps are sorted, I can use that code to generate (static?) RSS and ATOM feeds, and create page histories using diffs.
+It's all a bit too fragile, fix what I can.  Too many messes colliding.
+
+Flock it.
+
+Some sort of search system.
+
+A member system, and edit / manage system for their pages.
 
 Fix up linky conversion.  DONE, mostly.
 
-- This is the "page moved" problem, but now it's "page copied" and "page linked", a generic solution might work.
-- Rethink this, I might be going the wrong way.
 - Need to deal with real file name versus title.  Also symlink name not matching what it points to.
-- . Hmmm, might it be good policy to have a single canonical context for each page?  The "real URL" .md.md thing?
-- . Loop through the manually placed symlinks, adjusting "real URL" as we go.
-- + What to do about mulitple symlinks pointing to the same page?  Compare timestamps, most recent wins.
 
-Deal with complex directory trees.
+Use the default.template that comes with nYAW if none is availaable.
+
+"collapsing headings" I guess that means click on a heading to hide / show the content under that heading.
 
-- /testing/even should display as even/deeper on the testing page, coz even/ has no files, but even/deeper does.
-- On the other hand, I could just create an empty page with the sub directories shown as usual.
-- Scanning /usr/share/doc on my super desktop with looots of software installed will be fun.
-- On the gripping hand, this is where we want empty directories to vanish.
-- SOOOOOOO don't show empty directories, but show their subs that have content, even if that sub is deep, but as per usual stop once we find a sub on that branch.  EEEEW branches.
+- 🙈
+- Should do that for the main content and the menu TOC.
+- Also allow editing just a section, a section being the bits between one heading and the next.
+- + 📝 ✒️✏️🖊️🖋 🖌️🖍️
+- + When someone starts editing, create pagename_draft.md and .HTML, and update those when they want to see a preview.
+- + Or pagename-heading_draft.md
+- + Use that draft.md file as a lock on editing that file / section.
+- + Have the hourly cron job remove any stale ones.
 
+Add ATOM feed for single page.  Alas cgit only seems to have ATOM feed on the whole repo, not individual files.
+
+- However, once timestamps are sorted, I can use that code to generate RSS and ATOM feeds, and create page histories using diffs.
+- Instead of an hourly cron job to update everything, see if I can hook
+into Fos and Pm edit form's save function.  So when they save in the
+original wiki, the nyaw version gets updated, diffed, and ATOMed / RSSed.
+- + Actually a BFI method might do the trick, stick with doing a full scan 
+and only updating the stuff that needs it, likely that's only the page that just got saved.
+- git commit can have an arbitrary author / committer, so I can use that to make sure the person that actually made the change gets author / committer.
+- + There's also the email thing, but I don't want to spread email addresses around, not even sure we have them anyway.
+- Should also see if gitea can handle that, and if they want wiki content saved to git.
+
+Syntax highlighting in code blocks.  The highlight package looks promising, already had it installed on my desktop.
 
 ## Some ideas
 
-Mostly from something chomwitt deleted.
+Mostly from something chomwitt wrote -
 
-- "collapsing headings" I guess that means click on a heading to hide / show the content under that heading.
-- + Should do that for the main content and the menu TOC.
 - "validation", not sure exactly what that would validate.
 - I'll just quote some of the rest -
 
@@ -42,16 +59,8 @@ Mostly from something chomwitt deleted.
     ** for example search-completion to other workflows names-tags.
 ~~~
 
-- Syntax highlighting in code blocks.
+- Allow default.template files in sub folders.
 - Might be useful to automatically convert anything looking like a URL into a linky.
-- Allow default.template files in sub directories.
-
-Check the timestamps on the files, only update if source is newer than destination.  Meh, it's already 600 times faster than the pandoc version.
-
-- One quirk to watch for is if a URL path changes, the docs that have that URL need to be redone.
-- pandoc is a lot slower though, so do this for sure when dealing with that.
-- When scraping the web sites, they tend to be dynamically generated with no useful timestamp on them.
-- The web site scrape happens locally anyway, I can compare source file timestamps.
 
 Automate symlinks.
 
@@ -60,16 +69,28 @@ Automate symlinks.
 
 ## Try out
 
+lua-lpeg-patterns might be useful
+
+lua-luxio might be the wheel I'm reinventing?
+
+lua-wsapi-fcgi
+
 https://inclusive-components.design/tooltips-toggletips/ has some ideas about the "tooltip" HTML attribute, and what to do about it.  A rabbit hole that ends in javascript, but might be useful up to that point.
 
 htmx
 
 cgit has Lua
 
+lua-gall for git stuff
+
+lua-lace for access control
+
+lua-unbound
+
 
-## User system
+## Member system
 
-Reuse the user system from SledjChisl.
+Reuse the member system from SledjChisl.
 
 levels -
 
@@ -111,6 +132,6 @@ site, including configuration and modules.  Likely this is the person
 that set the system up in the first place.
 
 Admin should have access to everything that shell level has, but there's always things need tweaking at some lower level.
-Built in file browser might do the trick.  Would be useful for content creators to to organise the content.  Naturally should obey the permissions.
+Built in file browser might do the trick.  Would be useful for content creators to organise the content.  Naturally should obey the permissions.
 
 root level is whoever controls the server things are running on.  They can do anything at all.