Must be the irony of life that I was about to give up the TclCurl Debian package some time ago, and now I'm using it again for some very old and horrible web scraping code.

The world moved on to https but the Tcl http package only supports unencrypted http. You can combine it with the tls package as explained in the Wiki, but that seems to be overly complicated compared to just loading the TclCurl binding and moving on with something like this:

package require TclCurl
# download to a variable
curl::transfer -url https://sven.stormbind.net -bodyvar page
# or store it in a file
curl::transfer -url https://sven.stormbind.net -file page.html

Now the remaining problem is that the code is unmaintained upstream and there is one codebase on bitbucket and one on github. While I fed patches to the bitbucket repo and thus based the Debian package on that repo, the github repo diverted in a different direction.