M dot site load times suffer from the initial redirect, but generally are much smaller in request and KB count. RWD sites are the opposite: no redirect, but heavy on the requests and tonnage. That leads to a question that was discussed at a panel at Velocity Conference: Which is faster: M dot or RWD?
Now, there is a feature inside WPT to identify RWD sites in desktop, but it is not yet being recorded in the HTTPArchive. So I had to come up with criteria to categorize sites:
WHEN HOST(requests.url) LIKE 'm.%' then "M dot"
WHEN HOST(requests.url) LIKE 't.%' then "T dot"
WHEN HOST(requests.url) LIKE '%.mobi%' then "dot mobi"
WHEN HOST(requests.url) LIKE 'mobile%' then "mobile"
WHEN HOST(requests.url) LIKE 'iphone%' then "iphone"
WHEN HOST(requests.url) LIKE 'wap%' then "wap"
WHEN HOST(requests.url) LIKE 'mobil%' then "mobil"
WHEN HOST(requests.url) LIKE 'movil%' then "movil"
WHEN HOST(requests.url) LIKE 'touch%' then "touch"
WHEN HOST(requests.url) like host2 then "no change to domain"
WHEN HOST(requests.url) like domain2 then "drop www"
WHEN HOST(requests.url) not like domain2 then "new domain"
else "no redirect"
This finds the first HTML request, and looks to see if the URL shows any type of mobile specific redirect. I know I likely missed some of the mobile redirects, and will also miss sites that do a server side change (with no URL modification. I considered the first 9 categories of sites as m dot redirected sites.
To get the responsive list, I reached out to Guy Podjarny, who kindly shared the list of sites he discovered in December 2013. Yes, this list is a bit dated, but odds are that these sites have not abandoned their RWD sites (but I am likely missing RWD sites launched since then).
I limited my search to the top 1,000 mobile websites. Andy Davies ran these same top 1,000 websites using the Moto G devices in Dulles (running Kit Kat). I obtained similar results with both data sets:
About 10% of the sites are RWD, and 21-26% are m dot. As might be expected the m dot sites are generally smaller, both in number of number of requests and the total KB tonnages of the pages.
This of course leads us to which site type is faster? This data is from the Moto G dataset, as the iOS agents do not offer TTFB (but the other numbers are similar). As expected, the m dot sites are slower off the bat (TTFB) due to the redirect. In fact, TTFB is nearly 50% slower!
But what is really interesting is the idea that RWD sites are VERY competitive on Visually Complete and SpeedIndex scores. The median values are within 5% for both metrics. Even though it appears that RWD is faster, there is enough fluctuation in the data that we should probably call it a dead heat.
Digging into this even a bit more, it is evident by this graph that there are a number of very fast responsive websites being served to mobile handsets. The distribution of RWD appears to be faster than that of the m dot (albeit with a longer tail). In fact, if we omit sites that have SpeedIndex over 20,000 (9 m dot and 14 RWD sites), the median SpeedIndices drop to 7210.5 (RWD) and 9137 (m dot).
In conclusion, these findings show that RWD sites can indeed compete with m dot sites in terms of load time!
Thank you to @guypod abd @andydavies for the data and discussions at Velocity Conference!