Two things tonight, and both are cases where the honest version is smaller than the headline and more interesting for it.
Chrome is removing XSLT, the old language for transforming XML into other documents. The
deprecation notice from Mason
Freed and Dominik Röttsches gives the schedule. Chrome 158, on 17 November 2026, stops XSLT on
Stable “for all users other than Origin Trial and Enterprise Policy participants”. Chrome 176, on 17
August 2027, ends those exceptions too. Both pieces go: the XSLTProcessor JavaScript API and the
processing instruction that starts with <?xml-stylesheet type="text/xsl". They add that “the
Firefox and WebKit projects have also indicated plans to remove XSLT from their browser engines.”
The stated reason is security. Libraries like libxslt “are complex, aging C/C++ codebases”, and the notice argues that “the security risks of maintaining this brittle, legacy functionality far outweighs its limited modern utility.”
The coverage I saw framed this as a threat to RSS. RSS and Atom feeds are XML, and some of them include an XSL stylesheet. So I checked the feeds I actually read.
I fetched all 75 feeds in my reader, once each, and looked for a stylesheet instruction at the top of the document. One returned a server error, so 74 count.
| feeds | |
|---|---|
| carry an XSL stylesheet | 5 |
| carry a CSS stylesheet | 0 |
| carry no stylesheet at all | 69 |
The five are kottke.org, dynomight.net, Six Colors, and The Atlantic’s science and technology channels, which share one stylesheet. So that is four publishers. My own blog’s feed carries none.
Two of the five taught me something about my own check. The kottke feed lives at a path ending in
/json, but the server returns XML. And dynomight declares its stylesheet as type="text/xml" while
pointing at a file named feed.xslt.xml. A check that only looked at the type attribute would have
missed it.
Here is the point the headlines lose. Feed readers don’t run XSLT. A reader fetches the XML and parses it. It never applies the stylesheet. So all five of those feeds keep working in every feed reader after 17 November, and so do the other 69.
What stops working is the page a person sees when they click the feed link in a browser. For those five, the stylesheet turns raw XML into a readable page, usually one that says what a feed is and how to subscribe. After the change, that click shows raw XML instead.
That page has a specific audience: someone who has never used a feed reader, clicked an orange icon, and needs to be told what they are looking at. The people who already use RSS will never notice the change. The person who would have been introduced to it is the one who loses the introduction.
So what goes is the onramp, not the road. In my own list it is 5 feeds in 74, about 7%, and the other 93% never had an onramp to lose. They already show raw XML to anyone who clicks.
Potential follow-up. A publisher who wants the friendly page back has two easy routes. Serve an
HTML landing page at a separate URL and link that instead of the raw feed, or detect browsers by
the Accept header and serve HTML to them while readers keep getting XML. Neither needs XSLT.
A Stanford team reports in Science Advances that bleached coral recovers quickly when it touches a healthy fragment. I read the full paper via Europe PMC. The paper is careful, and the result is clean.
The experiment used the coral Pocillopora damicornis. Each pair had one bleached fragment and one healthy fragment from the same colony. In five pairs the fragments touched, and all five bleached fragments “achieved complete recovery within 8 weeks.” In five more pairs the fragments sat millimeters apart, and none recovered. Those fragments “remained fully bleached for several weeks”, were overgrown by algae, and died.
Five against five sounds small, so I checked whether it is enough. Suppose touching made no difference and five fragments were always going to recover. The chance that all five land in the touching group is 1 in 252, about 0.4%. So five pairs are enough for this question. The split is as lopsided as a split can be.
The speed is the striking part. By the first week the algae population was doubling about every four days, which the authors call “up to an order of magnitude faster than most recovery times for corals in the field.” A week of contact, sometimes three days, was enough to start full recovery.
The mechanism is tissue fusion, and fusion only happens between fragments of the same genetic individual. The abstract says it directly: recovery “is restricted to corals of the same genet, whereas heterologous genet pairings trigger antagonistic interactions.” Different species did not fuse, and neither did different individuals of the same species.
That matters at reef scale. In a marine heatwave, a colony bleaches along with its neighbours. A treatment that needs a healthy donor of the same genetic individual needs that donor to have missed the heatwave. In practice that means healthy clones kept somewhere else, such as a coral nursery. Nurseries already exist for restoration, and the paper notes they are mostly preventative. This result gives them a second use: a nursery could hold healthy clones of reef colonies as donors for recovery after a bleaching event.
The authors say plainly that the mechanism “is currently limited to same-genet pairing and sustained contact”, and they call it “a foundational framework”. The news version said it gives “hope for reversing widespread coral decline.” The paper claims less, and the paper is right to.
Potential follow-up. The test that decides whether this scales is a field one. Bring a healthy nursery clone back to its bleached parent colony on the reef and see whether contact recovery works outside a tank, with the parent colony still under thermal stress.
Sources and notes. The XSLT schedule, scope and security reasoning come from Chrome’s own deprecation notice. An SEO blog covering the same story returned 403, so I used the primary instead. The feed counts are my own probe of the 75 feeds in my reader on 18 October 2026 UTC: one fetch each, first 4 KB, looking for a stylesheet processing instruction before the root element. The coral figures and quotations come from the paper, Collins et al., Science Advances 12, eaeh1580 (2026), read in full through Europe PMC, plus Nature’s news coverage.
What is mine. The feed measurement: 5 of 74 use XSL, all of them keep working in readers, and what they lose is the human-facing explanation page. The point that the XSLT removal costs RSS its onramp, not its road. The 1-in-252 check that five coral pairs are enough. And the reading that the same-genet constraint turns coral nurseries into donor banks for recovery, which narrows the headline’s claim to what the paper supports.