alt Hacker NewsRemoving XSLT from browsers was long overdue and I'm saying that as ex-maintainer of libxslt who probably triggered (not caused) this removal. What's more interesting is that Chromium plans to switch to a Rust-based XML parser. Currently, they seem to favor xml-rs which only implements a subset of XML. So apparently, Google is willing to remove standards-compliant XML support as well. This is a lot more concerning.
Replies
https://issues.chromium.org/issues/451401343 tracks work needed in the upstream xml-rs repository, so it seems like the team is working on addressing issues that would affect standards compliance.
Disclaimer: I work on Chrome and have occasionally dabbled in libxml2/libxslt in the past, but I'm not directly involved in any of the current work.
> This is a lot more concerning.
I'm not so sure that's problematic. Probably browser just aren't a great platform for doing a lot of XML processing at this point.
Preserving the half implemented frozen state of the early 2000s really doesn't really serve anyone except those maintaining legacy applications from that era. I can see why they are pulling out complex C++ code related to all this.
It's the natural conclusion of XHTML being sidelined in favor of HTML 5 about 15-20 years ago. The whole web service bubble, bloated namespace processing, and all the other complexity that came with that just has a lot of gnarly libraries associated with it. The world kind of has moved on since then.
From a security point of view it's probably a good idea to reduce the attack surface a bit by moving to a Rust based implementation. What use cases remain for XML parsing in a browser if XSLT support is removed? I guess some parsing from javascript. In which case you could argue that the usual solution in the JS world of using polyfills and e.g. wasm libraries might provide a valid/good enough alternative or migration path.
> Removing XSLT from browsers was long overdue
> Google is willing to remove standards-compliant XML support as well.
> They're the same picture.
To spell it out, "if it's inconvenient, it goes", is something that the _owner_ does. The culture of the web was "the owners are those who run the web sites, the servants are the software that provides an entry point to the web (read or publish or both)". This kind of "well, it's dashed inconvenient to maintain a WASM layer for a dependency that is not safe to vendor any more as a C dependency" is not the kind of servant-oriented mentality that made the web great, not just as a platform to build on, but as a platform to emulate.
> Currently, they seem to favor xml-rs which only implements a subset of XML.
Which seems to be a sane decision given the XML language allows for data blow-ups[^0]. I'm not sure what specific subset of XML `xml-rs` implements, but to me it seems insane to fully implement XML because of this.
I think it might make more sense to use WebAssembly and make them as extensions which are included by default (many other things possibly should also be made as extensions rather than built-in functions). The same can be done for picture formats, etc. This would improve security while also improving the versatility (since you can replace parts of things), if the extension mechanism would have these capabilities.
(However, I also think that generally you should not require too many features, if it can be avoided, whether those features are JavaScripts, TLS, WebAssembly, CSS, and XSLT. However, they can be useful in many circumstances despite that.)
Given that you have experience working on libxslt, why do you think they should have removed the spec entirely rather than improving the current implementation or moving towards modern XSLT 3?
I was somewhat confused and irritated by the lack of a clear frontrunner crate for XML support in rust. I get that xml isn't sexy, but still.
> Currently, they seem to favor xml-rs which only implements a subset of XML.
What in particular do you find objectionable about this implementation? It's only claiming to be an XML parser, it isn't claiming to validate against a DTD or Schema.
The XML standard is very complex and broad, I would be surprised if anyone has implemented it in it's entirety beyond a company like Microsoft or Oracle. Even then I would question it.
At the end of the day, much of XML is hard if not impossible to use or maintain. A lot of it was defined without much thought given to practicality and for most developers they will never had to deal with a lot of it's eccentricities.
What's long overdue is them updating to a modern version of XSLT.
It’s interesting to see the casual slide of Google towards almost internet explorer 5.1 style behavior, where standards can just be ignored “because market share”.
Having flashbacks of “<!--[if IE 6]> <script src="fix-ie6.js"></script> <![endif]-->”