@MichaIng opened this Issue on April 24th 2021

Summary

With #16795 a robots.txt has been implemented which breaks tracking of bots. Tracking bots to derive the sites and frequency of search machine crawlers is valuable information, hence bots should by default be tracked like any other visitor. There is even a plugin available to track bots separately based on user agent, which totally lost its purpose with the robots.txt. I did not fully understand the reason for #16795, but it seems to be based on a single user having issues with google ads submission. However, in case of a single user, I'd say a custom solution makes more sense, than breaking a generally useful feature for all users with a file that is installed and re-created on every Matomo update, which hence cannot be avoided without regular manual interaction.

What makes sense, is preventing the Matomo login page from being indexed, as this is likely never meant to be public, but that is the case already via meta tag, which is generally the better way to do it (preventing crawlers from having to read an additional file every time): https://github.com/matomo-org/matomo/blob/4.x-dev/plugins/Login/templates/loginLayout.twig#L3-L5

In case of Google and Bing, crawlers are still not tracked after removing the robots.txt, failing on the tracker PHP request. But I didn't investigate this much further without having a confirmation first that tracking bots is generally a wanted feature, which in terms means to remove the robots.txt that breaks it.

@tsteur commented on April 26th 2021 Member

Hi @MichaIng it did affect actually quite a few users which is why it was added. Generally though, there's a wildcard configured for the tracking endpoints: https://github.com/matomo-org/matomo/blob/4.3.0-b3/robots.txt#L30-L32 and things should work (in theory). Is it maybe that you use a different endpoint for tracking? I see we could for example also add js/index.php and js/tracker.php.

@MichaIng commented on April 26th 2021

The wildcard + Disallow means that all clients who respect the robots.txt will not read the endpoints.

The idea of the robots.txt text was that the tracking endpoints shall be blocked for all bots (hence the wildcard) while only known search machines crawlers shall not crawl any of the contained files. This was required as otherwise wget fails to access anything (e.g. reports) as well.

The following would be required to all access to tracking endpoints while blocking access for search engine crawlers to everything else.

User-agent: *
Allow: /matomo.php
Allow: /piwik.php

But that would still now be sufficient, as in the first place piwik.js/matomo.js need to be allowed and all other files that might be indirectly invoked by those scripts:

User-agent: *
Allow: /matomo.js
Allow: /matomo.php
Allow: /piwik.js
Allow: /piwik.php
@tsteur commented on April 26th 2021 Member

Thanks for the reply @MichaIng . I was reading the htaccess wrong and it is blocking all bots. I'll try follow up with this internally next week or the week after and get back to you. We may need to have some kind of option for this and re-evaluate whether it actually helped for Google ads submission or not.

@tsteur commented on April 26th 2021 Member

Moving this for now into 4.3 milestone but there's nothing to be done just yet. We need to discuss what could be options to make this work again.

@tsteur commented on May 9th 2021 Member

We would for now remove the rules for matomo.php and piwik.php and see if it causes issues again re Google.

@MichaIng commented on May 10th 2021

Many thanks. matomo.js and piwik.js need to be allowed as well, AFAIK. Will test it.

@tsteur commented on May 10th 2021 Member

@MichaIng there shouldn't be a need for the bots to access these

@mattab commented on May 10th 2021 Member

@tsteur I think the .js files are needed for the bots that execute JavaScript if you want them to also execute Matomo JavaScript and track them...

@MichaIng commented on May 10th 2021

But the JavaScript is what is added to the document via tracking code. And the JavaScript then performs the actual tracker call:

<!-- Matomo -->
<script type="text/javascript">
  var _paq = window._paq = window._paq || [];
  /* tracker methods like "setCustomDimension" should be called before "trackPageView" */
  _paq.push(['trackPageView']);
  _paq.push(['enableLinkTracking']);
  (function() {
    var u="https://domain.com/matomo/";
    _paq.push(['setTrackerUrl', u+'matomo.php']);
    _paq.push(['setSiteId', '1']);
    var d=document, g=d.createElement('script'), s=d.getElementsByTagName('script')[0];
    g.type='text/javascript'; g.async=true; g.src=u+'matomo.js'; s.parentNode.insertBefore(g,s);
  })();
</script>
<!-- End Matomo Code -->

Results in:

<script async src="https://domain.com/matomo/matomo.js"></script>

Which performs the matomo.php call, or do I misunderstand the logic?

Now I wonder why the short code snippet does not do the matomo.php call directly, like it is for non-JS clients:

<noscript><p><img src="https://domain.com/matomo/matomo.php?idsite=1&amp;rec=1" style="border:0;" alt="" /></p></noscript>

But at least the Google crawler loads scripts. If blocked by the robots.txt, the mobile-friendly test does not show a matomo.php access error anymore but a matomo.js error instead, which indicates that it is loaded first.

@tsteur commented on May 10th 2021 Member

@tsteur I think the .js files are needed for the bots that execute JavaScript if you want them to also execute Matomo JavaScript and track them...

@mattab that could be actually a good thing as then also the accessible matomo.php wouldn't be an issue maybe. it be just like someone with an ad blocker etc.

We can add it if really needed but I reckon it could be a good compromise between making tracking bots work from the importer script and not having the crawlers access the JS. But we can add it for people that want to track bots without the importer.

@mattab commented on May 11th 2021 Member

Well, as a workaround, people could always go in their robots.txt and manually delete the matomo|piwik.js lines? So i suppose it's fine this way as well

@MichaIng commented on May 11th 2021

tracking bots work from the importer script

You mean calling matomo.php directly from the HTML <script> snippet, like the <noscript> variant does? If so, how would that look like (as the admin interface only suggests the above pasted variant)? I recognised that the matomo.php call from matomo.js has a long query string which passes client info, while the suggested <noscript> call passes the site ID only. Does matomo.php gather client info then from what the webserver passes or is the info not complete?

Well, as a workaround, people could always go in their robots.txt and manually delete the matomo|piwik.js lines?

Definitely, but it isn't great when one needs to do this manually after every Matomo update. I apply beta versions as well, that sums up to regular SSH access to remove/alter that robots.txt to not cause contradicting info for crawlers. So if we find a way to prevent crawlers from indexing the Matomo login page without breaking either bot tracking or Google ads submission, that would be awesome.

Btw, could someone give me a link regarding the Google ads submission issue? I didn't understand yet how blocking bot access can have an effect there. The linked issues I found are about preventing search engines from indexing the Matomo login page, which is totally reasonable but would require to block index.php only (AFAIK, as non-HTML resources are not indexed anyway). All sub directories have a .htaccess to prevent access, so that is covered as well.

@tsteur commented on May 11th 2021 Member

@MichaIng we've added the tracker files for now to the allow list so it just all just work.

@MichaIng commented on May 11th 2021

Many thanks! I'll give it a try.

Btw, the long user agent list is still required to allow e.g. CSV report downloads via wget, right? I wonder if it would be easier (and safer) to apply the rules to all User-agent: * and then grant report access to wget specifically? Not sure if there are other download tools which respect robots.txt, but there are definitely a lot more crawler/bot user agents, hence the list is difficult to maintain.

@tsteur commented on May 11th 2021 Member

@MichaIng give it a go and let us know if it doesn't work. wget user agent should not be blocked.

This Issue was closed on May 10th 2021
Powered by GitHub Issue Mirror