Allen indicated 4 problems, remailing here so it is visible in the mailing list.
He seems to be willing to help, but the size of Nar web is a problem.
I'd be willing to consider it if I were more familiar with the internals of GRAMPS, but it would take time to get up to speed on it. After all, NarrativeWeb.py is almost 8500 lines long.
I've identified at least four problems that need to be fixed:
1) Descriptions have been removed from event tables. It is unclear to me why this would have been done unless it was to encourage use of the event pages.
2) Media objects associated only with family events are omitted from the web site if family pages are generated.
3) The references section on media pages is almost worthless if event pages are generated (the references are to the events and each is labeled "Unknown").
4) Media objects associated only with sources are omitted from the web site.
I suspect it might be possible to fix item 1 by making a few changes to the methods event_header_row and display_event_row, and this is a serious problem in my opinion.
Family pages were new to 3.4 and I suspect development of them wasn't completed in time for release. A work-around to those problems is to not generate family pages, which is what I've elected to do for now.
I've also elected not to generate event pages on my web sites because of item 3 and because I don't find them all that useful.
Item 4 is one which I would very much like to see fixed, but I suspect it is due to a circular dependency and that it would require significant restructuring of the code to fix it.
I thought I also read on another list that gramps-connect might have some impact on the future of the narrative web report and that someone might be working on better use of CSS files to reduce the amount of generated HTML.
I do have over 20 years of programming experience (mostly C and C++ for scientific/engineering applications) and I've been using Python for about six years, but most of the Python programs I've written have been for performing data analysis on large sets of numerical data, much different than what you are doing in GRAMPS.