[wp-trac] [WordPress Trac] #56990: Render blocking CSS `classic-themes.css` unnecessarily enqueued

[WordPress Trac]

#56990: Render blocking CSS `classic-themes.css` unnecessarily enqueued
------------------------------------------+-------------------------------
 Reporter:  adamsilverstein               |       Owner:  flixos90
     Type:  defect (bug)                  |      Status:  assigned
 Priority:  normal                        |   Milestone:  6.3
Component:  Themes                        |     Version:  6.1
 Severity:  normal                        |  Resolution:
 Keywords:  needs-patch needs-unit-tests  |     Focuses:  css, performance
------------------------------------------+-------------------------------
Changes (by flixos90):

 * keywords:  has-patch needs-testing => needs-patch needs-unit-tests
 * milestone:  6.2 => 6.3


Comment:

 I have conducted another performance benchmark for this locally, using
 [https://github.com/GoogleChromeLabs/wpp-research/tree/add/benchmark-web-
 vitals/cli#benchmark-web-vitals this CLI script] to measure FCP, LCP, and
 TTFB. Running locally is better for benchmarks than using WebPageTest (or
 generally requests to external sites) due to the higher variance in those
 scenarios.

 I ran 20 requests each to WP 6.2 Beta 2 with vs without the change from
 https://github.com/WordPress/wordpress-develop/pull/4012 applied.

 The results were inconclusive, so then I thought it may be useful to add
 another metric "LCP - TTFB", essentially compute that for every request
 and get the median (which, just to clarify, is not the same as subtracting
 median TTFB from median LCP!). This metric is more effective to assess the
 performance impact here, since the server response time (TTFB) should be
 ignored for this change, as what we are changing affects only client-side
 performance, if anything.

 The additional metric did nothing but confirm that there is really no
 clear benefit from either approach. Sometimes the inline style was faster,
 other times the external file. See the data for "LCP - TTFB" below.

 1. Comparison (20 requests each): 133ms (inline style) vs 137.5ms
 (external stylesheet)
 2. Comparison (20 requests each): 134.65ms (inline style) vs 130.65ms
 (external stylesheet)

 This additional benchmark confirms what was mentioned above: It doesn't
 really make a difference whether we have an inline style or external
 stylesheet for this. For that reason, what remains here is a more
 appropriate fix, which would be to somehow only load this stylesheet when
 at least one of those blocks is used on the page in a classic theme. This
 is a whole lot more complex to figure out though (if even reasonably
 possible), so I'm removing this from the 6.2 milestone as such a change
 would have the chance of introducing other problems, which we are too late
 for now.

-- 
Ticket URL: <https://core.trac.wordpress.org/ticket/56990#comment:42>
WordPress Trac <https://core.trac.wordpress.org/>
WordPress publishing platform