Experiment: Run the Tag Processor on every tag on output.

[dmsnell]

Using this to get a baseline measurement on a kind of "worst case" scenario with the HTML Tag Processor: how much delay do we incur if we're running the processor over the final page output and processing every tag?

Why would this be useful information or a useful practice?

• a new set of WordPress filters to adjust attribute values more robustly
• look for security issues in URLs sent to the browser
• look for accessibility issues
• append extras like srcset values in one final go
• check for duplicate id values

In this experiment every src and href attribute value is parsed.

Initial results

• very promising: no measurable impact on the page render for twentytwentyfour's homepage
• on my laptop it's processing at around 100 MB/s for the twentytwentyfour homepage, the speed improvement probably due to the large inline stylesheet on the page

From my estimates this should be adding around at least 1ms of page render time, but we aren't able to statistically infer that from the results. Why this is I'm not sure because I would expect a 1% change to certainly appear, but something else might be masking the impact.