5 ms·

The survey uses multiple variables to figure out server side language.

by tored 5y ago

theamk 5y ago
Sure, but this is still going to be pretty limited when detecting non-PHP languages. How are you going to detect a language which puts no extra headers in HTTP response and has no associated extensions? But, luckily for us, w3techs lets you check individual websites... and the data is often wrong or missing github on server side is Javascript! https://w3techs.com/sites/info/github.com https://w3techs.com/sites/info/github.com (it is actually Ruby) Djagoproject.com on server side is unknown! https://w3techs.com/sites/info/djangoproject.com https://w3techs.com/sites/info/djangoproject.com (it is actually Django, which is a Python web framework: https://github.com/django/djangoproject.com https://github.com/django/djangoproject.com) microsoft.com was using PHP and WordPress "unil recently" https://w3techs.com/sites/info/microsoft.com https://w3techs.com/sites/info/microsoft.com gitlab.com is "ASP.NET, Python and Ruby (on subdomain) https://w3techs.com/sites/info/gitlab.com https://w3techs.com/sites/info/gitlab.com google.com, youtube.com and amazon.com are unknown https://w3techs.com/sites/info/google.com https://w3techs.com/sites/info/google.com etc.. ======== I stand by my opinion that w3techs is junk when determining types of server-side language. The only reason there are so much PHP is because PHP is the biggest major language left which is easy to detect. And they hide the low quality of the results it by not including "unknown" category in the graph -- with that category, it would be much clearer how bad their detection works.
- tored 5y ago
  You are free to think that w3techs is junk, but you initial claim was wrong. https://w3techs.com/disclaimer https://w3techs.com/disclaimer
  - theamk 5y ago
    I think I got most of this correct? My initial claims were: "bad reporting" -- I this is fully justified. The arstechnica report says things like "PHP [...] holds a 78.9 percent share as of today" without mentioning that this is not share of all servers, this is a share of servers where w3techs algorithm works. Which is a pretty low share based on my spot check. a website quote saying the percentages are not from all websites, but only from one where they know programming languages -- this is correct quote, see https://w3techs.com/technologies/details/pl-php https://w3techs.com/technologies/details/pl-php . "graph only covers the languages which leak their names into headers" -> I agree, this is wrong, I should have omitted "into headers" part. But I think it is a pretty minor difference and it does not change the other points. "The “unknown” category is missing completely," -- this is correct as well. Do you see "unknown server-side language" category in the graph or anywhere else in the report? Overall, that arstechnica article and that report are both omitting a crucial piece of information which makes the data unreliable.
    - tored 5y ago
      Ars technica's article is wrong, that is true. It claims that the survey says that PHP has a lead in server-side programming languages, that is not what the survey says, the survey only measures websites, not server-side programming languages in general. What I meant was that your original claim that the survey measures response headers only, that is false.
      - theamk 5y ago
        > the survey only measures websites, not server-side programming languages in general. the survey only measures websites where w3tech's algorithm works Or, in other words, "PHP maintains an enormous lead in the websites on which w3tech's scanner returned something" does not sound that impressive, does it?
        tored 5y ago
        I guess it depends on the failure rate, but the thing is that PHP has an overwhelming lead over the others that is not even close even if you would combine everyone else vs PHP, can the failure rate be that big? What this tells us is that still today PHP is really good fit for websites, but language makers seems to focus on solving other non website related problems but then hype their technologies that it can be used for websites when few of the web developers agree. That is a problem. Edit: And then we have the reoccurring PHP hate campaigns, and it seems to not only be something online but also within academia, which of course creates an illusion that PHP is not useful when we actually need more PHP developers.
        theamk 5y ago
        > can the failure rate be that big? yes? have you seen my post? https://news.ycombinator.com/item?id=28538595 https://news.ycombinator.com/item?id=28538595 out of 6 major websites I thought of (gitlab, github, microsoft, youtube, amazon), it got all of them wrong. I dug a bit further and while I could not find the recent data, I found some details: - For Jan 2013, Netcraft said "244M sites" running PHP: 244M sites: https://www.php.net/usage.php https://www.php.net/usage.php - For Jan 2013, Netcraft said they had 630M sites total: https://news.netcraft.com/archives/2013/01/07/january-2013-web-server-survey-2.html https://news.netcraft.com/archives/2013/01/07/january-2013-w... - That means that according to Netcraft, in Jan 2013, 39% of sites were running PHP - For the same date, w3techs were showing 75% of websites their scanner works on were running PHP This means that w3techs can be easily more than 100% off -- and I don't think it got better over time. I bet if we saw "unknown" figure, then PHP would get less than 50% of the sites, and dropping fast.

The survey uses multiple variables to figure out server side language.

by tored 5y ago

The survey uses multiple variables to figure out server side language.

Sure, but this is still going to be pretty limited when detecting non-PHP languages. How are you going to detect a language which puts no extra headers in HTTP response and has no associated extensions? But, luckily for us, w3techs lets you check individual websites... and the data is often wrong or missing github on server side is Javascript! https://w3techs.com/sites/info/github.com https://w3techs.com/sites/info/github.com (it is actually Ruby) Djagoproject.com on server side is unknown! https://w3techs.com/sites/info/djangoproject.com https://w3techs.com/sites/info/djangoproject.com (it is actually Django, which is a Python web framework: https://github.com/django/djangoproject.com https://github.com/django/djangoproject.com) microsoft.com was using PHP and WordPress "unil recently" https://w3techs.com/sites/info/microsoft.com https://w3techs.com/sites/info/microsoft.com gitlab.com is "ASP.NET, Python and Ruby (on subdomain) https://w3techs.com/sites/info/gitlab.com https://w3techs.com/sites/info/gitlab.com google.com, youtube.com and amazon.com are unknown https://w3techs.com/sites/info/google.com https://w3techs.com/sites/info/google.com etc.. ======== I stand by my opinion that w3techs is junk when determining types of server-side language. The only reason there are so much PHP is because PHP is the biggest major language left which is easy to detect. And they hide the low quality of the results it by not including "unknown" category in the graph -- with that category, it would be much clearer how bad their detection works.

tored 5y ago
You are free to think that w3techs is junk, but you initial claim was wrong. https://w3techs.com/disclaimer https://w3techs.com/disclaimer
- theamk 5y ago
  I think I got most of this correct? My initial claims were: "bad reporting" -- I this is fully justified. The arstechnica report says things like "PHP [...] holds a 78.9 percent share as of today" without mentioning that this is not share of all servers, this is a share of servers where w3techs algorithm works. Which is a pretty low share based on my spot check. a website quote saying the percentages are not from all websites, but only from one where they know programming languages -- this is correct quote, see https://w3techs.com/technologies/details/pl-php https://w3techs.com/technologies/details/pl-php . "graph only covers the languages which leak their names into headers" -> I agree, this is wrong, I should have omitted "into headers" part. But I think it is a pretty minor difference and it does not change the other points. "The “unknown” category is missing completely," -- this is correct as well. Do you see "unknown server-side language" category in the graph or anywhere else in the report? Overall, that arstechnica article and that report are both omitting a crucial piece of information which makes the data unreliable.
  - tored 5y ago
    Ars technica's article is wrong, that is true. It claims that the survey says that PHP has a lead in server-side programming languages, that is not what the survey says, the survey only measures websites, not server-side programming languages in general. What I meant was that your original claim that the survey measures response headers only, that is false.
    - theamk 5y ago
      > the survey only measures websites, not server-side programming languages in general. the survey only measures websites where w3tech's algorithm works Or, in other words, "PHP maintains an enormous lead in the websites on which w3tech's scanner returned something" does not sound that impressive, does it?
      - tored 5y ago
        I guess it depends on the failure rate, but the thing is that PHP has an overwhelming lead over the others that is not even close even if you would combine everyone else vs PHP, can the failure rate be that big? What this tells us is that still today PHP is really good fit for websites, but language makers seems to focus on solving other non website related problems but then hype their technologies that it can be used for websites when few of the web developers agree. That is a problem. Edit: And then we have the reoccurring PHP hate campaigns, and it seems to not only be something online but also within academia, which of course creates an illusion that PHP is not useful when we actually need more PHP developers.
        theamk 5y ago
        > can the failure rate be that big? yes? have you seen my post? https://news.ycombinator.com/item?id=28538595 https://news.ycombinator.com/item?id=28538595 out of 6 major websites I thought of (gitlab, github, microsoft, youtube, amazon), it got all of them wrong. I dug a bit further and while I could not find the recent data, I found some details: - For Jan 2013, Netcraft said "244M sites" running PHP: 244M sites: https://www.php.net/usage.php https://www.php.net/usage.php - For Jan 2013, Netcraft said they had 630M sites total: https://news.netcraft.com/archives/2013/01/07/january-2013-web-server-survey-2.html https://news.netcraft.com/archives/2013/01/07/january-2013-w... - That means that according to Netcraft, in Jan 2013, 39% of sites were running PHP - For the same date, w3techs were showing 75% of websites their scanner works on were running PHP This means that w3techs can be easily more than 100% off -- and I don't think it got better over time. I bet if we saw "unknown" figure, then PHP would get less than 50% of the sites, and dropping fast.