Mình đang xây dựng FeedBase – công cụ đơn giản để tập hợp và sử dụng hiệu quả phản hồi sản phẩm từ người dùng & đội nhóm. Thay vì feedback bị rải rác ở email, Slack, ticket..., FeedBase giúp gom lại, cho vote, bình luận và xác định ưu tiên dễ dàng. Thiết kế nhẹ, dễ tích hợp, phù hợp startup và SaaS quy mô nhỏ. Nếu bạn đang loay hoay với đống ý kiến lộn xộn, hãy thử và cho mình xin feedback nhé! #SideProject #ProductFeedback #FeedBase #StartupTool #CôngCụSảnPhẩm #PhảnHồiNgườiDùng #SaaS #ProductMa

Feedbase year 9

Looking at #feedbase here, it looks like I forgot to write about the 8th year of Feedbase last year.

Let's forget about that and write about the 9th year then. I'm still using Feedbase as a convenient means to keep up with RSS feeds ranging from news and podcasts, to Youtube channels, status information and what have you.

Rachel by the Bay created a "Feed Reader Behaviour Project" a while back, which nudged me to add a last_fetch timestamp to the groups, and skip fetching a group if it was less than an hour since last time.

This summer I also switched the website from the library Spock to twain instead, as it was easier when upgrading to Debian 13 (trixie), and I had already switched my other websites to twain.

Oh, and I have added a way to "properly" rename groups, where the original name stays around - but is filled with artificial articles all telling you to subscribe the the new name.

The number of commits added up to 34 in 2025, up from 21 in 2024, and down from 51 in 2023. I still haven't cleaned up the code and released it for public consumption.

The development in connections over the years:

2021 2022 2023 2024 2025 Unique IPs 479 506 1358 2859 3093 IPv4 324 369 1126 1753 2291 IPv6 155 137 232 1106 802 Only once 138 225 483 1224 1224

Quite spooky that the number of IPs only connecting once was exactly the same in 2024 and 2025!

Interestingly, I have not been the one connecting the most often since 2022 - I am down to 7th in 2025! The 6 most connecting IPs have connected more than a thousand times in the past year.

The total number of articles is nearing the 10 million mark.

Filtering on Supersedes header

Recently the question about how to filter (score) on the Supersedes: header when reading news via NNTP in Gnus was raised.

This is especially relevant on Feedbase, where updates to RSS/Atom entries are recorded as new articles that supersede previous ones. If you don't care about updates, down-scoring articles with Supersedes: is a nice solution.

Newsreaders, such as Gnus, can only do that effectively on the fields returned by the XOVER command in NNTP, so while somebody figured out how to do it by making Gnus fetch headers for all articles, it was slow.

The LIST OVERVIEW.FMT command in NNTP gives a list of the fields returned by XOVER, so I thought an easy solution would be to just add Supersedes: to that list, and to the output of XOVER.

After having done that, I still wasn't able to get Gnus to score on the new field.

Tracing through the Emacs lisp code with M-x edebug-defun didn't make me any wiser to the problem, it looked like it was comparing with "nil" rather than the value I knew XOVER was returning.

The next day I tried tracing some more, and finally realized that I was missing the name of the header in the information I added to XOVER!

After fixing that, this scoring rule worked as expected:

(("extra" ("." -1000 nil r "Supersedes")))

\o/