Super slow recursive download when using filter

I might being doing something wrong here, but it seems wget2 lacks some way to download just the file specified in --accept argument

Fast -> Using old wget: it downloads all index* files but then starts downloading the file specified in --accept argument straightaway:
wget -r -nd --no-parent -w 2 http://slackware.uk/slackware/slackware64-current/slackware64/a/ --accept=kernel-firmware-*.txz

Slow -> Using new wget2: it tries to download all files alphabetically until it reaches the file specified in --accept argument:
wget2 -r -nd --no-parent -w 2 http://slackware.uk/slackware/slackware64-current/slackware64/a/ --accept=kernel-firmware-*.txz

Using wget2 2.2.0

changed the description

There seem to be two issues.

1. the fnmatch only matches if you use --accept=*/kernel-firmware-*.txz (this needs to be amended to be wget1 compatible).

2. Wget2 is very thorough - despite of --accept, it queries all found files whether they are CSS or HTML or atom/rss. This requires a HEAD request to be made to figure out the mime type. The old wget1 just looks at the filename extension (.html or .css). If you want to fnmatch without a HEAD request, use --filter-urls.

While 1 needs to be checked, 2 potentially looks much deeper into websites to find your matching files. The downside is that it will take longer and more needs requests (mostly HEAD requests).