Repository navigation
Conversation
/rss.xml has 404'd since the Astro upgrade: the endpoint exported `get`, and Astro now only calls `GET`. The feed is rebuilt on a shared helper that skips drafts, sorts newest first, and links to /blog/<slug> without the trailing slash @astrojs/rss adds by default (posts are built as blog/<slug>.html). Each product tag also gets its own feed at /rss/<product>.xml, so CheerpJ users can follow /rss/cheerpj.xml. The blog index lists every feed, and each post lists the main feed plus its products' feeds, as <link rel="alternate"> tags so feed readers find them. Shell gains a named "head" slot for that. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Deploying labs-browserpod-previews with
|
| Latest commit: |
f73dd9e
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://e07f1fa5.labs-browserpod-previews.pages.dev |
| Branch Preview URL: | https://fix-rss.labs-browserpod-previews.pages.dev |
Deploying labs-cheerpj-previews with
|
| Latest commit: |
f73dd9e
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://982de613.labs-cheerpj-previews.pages.dev |
| Branch Preview URL: | https://fix-rss.labs-cheerpj-previews.pages.dev |
GabrielaReyna
left a comment
GabrielaReyna
left a comment
There was a problem hiding this comment.
No blockers and no bugs I can find, the suggestions below are polish:
- Nobody can find the feeds yet. The
<link rel="alternate">tags work for bots, but browsers stopped surfacing them years ago, so a human on the blog page has no way to know a feed exists. A visible subscribe link in the filter row would fix it, the component already knows which product is selected. - Each feed should state its own address (
<atom:link rel="self">). Costs one line, and it’s what keeps/rss/cheerp.xmland/rss/cheerpj.xmldistinct once a bot picks them up second-hand. - An RSS icon in the footer, so the feed is reachable from any page and not just the blog. Fine as a follow-up change.
- Category feeds can wait. The blog filters by product and category, but we only have product feeds. I’d rather check the
/rss.xmlaccess logs in a month before building more.
Others:
Minor: the item counts in the description look one off. I get 50 and 17, not 51 and 18.
| description: product | ||
| ? `Releases, technical deep dives and news about ${product} from the Leaning Technologies engineering blog.` | ||
| : SITE_DESCRIPTION, | ||
| site: context.site!, |
There was a problem hiding this comment.
Now that we have five feeds instead of one, could we add a few bits of channel metadata? They’re what aggregators and feed readers look for, and they all go in the same rss() call.
Right now only the <link> tag exists, and it points at the site root. So if you open rss/cheerpj.xml today, nothing inside tells you you’re looking at the CheerpJ feed rather than the Cheerp one or the main one.
The fix is an <atom:link rel="self"> tag, which points at the feed file itself, so each feed states its own address. It matters when a feed gets copied, mirrored, or pulled through a proxy: the bot can still find its way back to the original. And it matters for us specifically, because five feeds in the same folder look very similar.This is what keeps /rss/cheerp.xml and /rss/cheerpj.xmlclearly distinct downstream.
Worth adding <language> and <lastBuildDate> while we’re in there too. lastBuildDate should be the newest post’s date rather than the build time, otherwise every deploy makes the feed look changed when nothing was added.
| { | ||
| [undefined, ...FEED_PRODUCTS].map((product) => ( | ||
| <link | ||
| slot="head" | ||
| rel="alternate" | ||
| type="application/rss+xml" | ||
| title={feedTitle(product)} | ||
| href={feedPath(product)} | ||
| /> | ||
| )) | ||
| } |
There was a problem hiding this comment.
Right now the <link> tags are the only way to find these feeds, and Chrome and Safari stopped surfacing those years ago. So in practice a reader on this page has no way to discover we have a feed. Can we add a visible subscribe link?
| : post.data.title, | ||
| description: post.data.description, | ||
| pubDate: post.data.pubDate, | ||
| link: `/blog/${post.slug}`, |
There was a problem hiding this comment.
Minor heads up: dropping the trailing slash changes every item’s guid. Unsure if this hasn't been 404 forever, but if not it might notify 50 new posts?
https://labs.leaningtech.com/rss.xmlreturns 404. The endpoint exportedget, and since the Astro upgrade onlyGETis called. The build warned about it:No API Route handler exists for the method "GET" for the route "/rss.xml".What changes
/rss.xmlworks again. It's built by a shared helper (sites/labs/src/lib/feeds.ts) that leaves out drafts, sorts newest first and passes explicit fields: title (with subtitle), description, date, and tags and categories as<category>. The old code spread all of the frontmatter into each item, including the hero image object./rss/cheerpj.xml,/rss/cheerp.xml,/rss/cheerpx.xml,/rss/browserpod.xml, filtered by the post'stags.@astrojs/rssadds one by default, but posts are built asblog/<slug>.html, so each feed link would have cost a redirect.<link rel="alternate">, and each post lists the main feed plus its products' feeds.Shell.astrogets a namedheadslot for this. It's empty wherever it isn't used.Checked locally
astro checkreports 0 errors, and Prettier passes.rss.xmlhas 51 items,rss/cheerpj.xmlhas 18.dist/.rss.xmlandrss/cheerpj.xml. It isn't part of this PR.After deploy
https://labs.leaningtech.com/rss.xmlandhttps://labs.leaningtech.com/rss/cheerpj.xmlanswer 200 withapplication/xmlorapplication/rss+xml.https://labs.leaningtech.com/blogoffers the five feeds.🤖 Generated with Claude Code