Should Archive.org Ignore Robots.txt Directives And Cache Everything? (archive.org)

Archive.org argues robots.txt files are geared toward search engines, and now plans instead to represent the web "as it really was, and is, from a user's perspective." We have also seen an upsurge of the use of robots.txt files to remove entire domains from search engines when they transition from a live web site into a parked domain, which has historically also removed the entire domain from view in the Wayback Machine... We receive inquiries and complaints on these "disappeared" sites almost daily."
In response, Slashdot reader Lauren Weinstein writes: We can stipulate at the outset that the venerable Internet Archive and its associated systems like Wayback Machine have done a lot of good for many years -- for example by providing chronological archives of websites who have chosen to participate in their efforts. But now, it appears that the Internet Archive has joined the dark side of the Internet, by announcing that they will no longer honor the access control requests of any websites.
He's wondering what will happen when "a flood of other players decide that they must emulate the Internet Archive's dismal reasoning to remain competitive," adding that if sys-admins start blocking spiders with web server configuration directives, other unrelated sites could become "collateral damage."

But BoingBoing is calling it "an excellent decision... a splendid reminder that nothing published on the web is ever meaningfully private, and will always go on your permanent record." So what do Slashdot's readers think? Should Archive.org ignore robots.txt directives and cache everything?

Comments

yeah

[Anonymous Coward]

yeah!

Re:yeah

[ArmoredDragon]

Law of headlines indeed, and there's already an established way for web developers to indicate that they don't want content cached or archived while still being searchable:

<meta name="robots" content="noarchive">

So archive.org could just honor that, and the problem would be solved. Google honors exactly this.

Re:yeah

[Zocalo]

Even more specific robots.txt directive for this instance:

User Agent: ia_archiver
Disallow: /

As is often the case, Lauren is going off half-cocked with only part of the story. The IA already has a policy for removal requests (email info@) and is only considering expanding their current position of ignoring robots.txt on sites outside their current "test zone" of the .gov and .mil gTLD domains and have not had any problems. They probably will do that (and for their archival purposes it's a good idea in principle), but I think it's only fair to see whether or not they listen to the feedback and provide some specific opt-out policy and technical mechanisms like at least honoring either of the above prior to going live on the rest of the Internet before starting to scream and shout. It's going to be a two-way street anyway because they're going to find a lot more sites that feed multiple-MB of pseudo-random crap to spiders that ignore robots.txt to try and do things like poison spammer's address lists, so it's actually in their best interests to provide an opt-out they honor.

Besides, it's going to be interesting to see what kind of idiotic crap web admins who should know better think is safely hidden and/or secured because of robots.txt - it's useful to know who is particularly clueless so you can avoid them at all costs. :)

Re:yeah

[Zocalo]

IA does still spider, but they seem to use a more nuanced system than the rudimentary "start at /, then recursively follow every link" approach used by more trivial site spider algorithms. Firstly, they don't download an entire site in one go - they spread things out over time to avoid putting large spikes into the traffic pattern which is more friendly for sites that are bandwidth limited and on things like "xGB/month" plans. Secondly, they have a "popularity weighting" system that governs the order they spider and refresh sections of a given site, which is the main reason for the difference between the level of content for popular and less popular sites - although I have no idea whether that's based entirely off something like the site's Alexa ranking or is also weighted against how dynamic the content is (e.g a highly dynamic site like Slashdot would get a bump up the priority, whereas a mostly static reference site might get downgraded). Combine the two approaches and you get the results you are seeing: major web homepages get spidered more or less every day with several levels of links retrieved, while some random personal blog only get spidered every few weeks or more, and only with the homepage and first level or two of links ever getting looked at.

Random generated content

[DrYak]

It is also for variable random content. Imagine a service that returns a webpage containing the product (of the multiplication) of two numbers, followed by a list of links to ten other random number pairs you could try. It would take a 1kB page to write, but infinite space to archive *all* the results

And archive.org already has a correct behaviour for that :
- it wont try to download all infinity of solution in one go (e.g.: generating giga-byte worth of data out of the 1kB Perl/PHP/NodeJS/whatever source)
- instead it will occasionally rescan the page, every few days (more or less frequently, depending on popularity of the links)
It provides a small glimpse of what a user could have seen back then on the website.

By the way, back in the 2000s, this was exactly a popular way to poison SPAM robots spiders who where scanning the web for e-mail addresses.
- Either they honour robots and not scan that or any other sources of e-mail on the site.
- Or they attempt to ignore robots.txt and follow links they aren't authorised to, and end-up siphonning giga-bytes worth bogus e-mails addresses auto-generated by small perl script, which will pollute their base of harvested addresses.

Archive.org's spider might by a tiny bit more susceptible to this kind of things.
Bot as much as a SPAM email-harvesting spider (which will try to download as much as possible, much more aggressively than archive.org), but still such a labyrinth of links might get archive lost.

I conducted a 2 yr experiment on Internet Archive

[Anonymous Coward]

I wanted to know if it was possible to delete content from the Internet Archive. Their FAQ and support staff were very vague and only referred me to the robots.txt file. I found that they archive everything even if you tell them not to. The robots.txt file only controls whether or not the public can view it.

Experiment 1) Buy an expired domain and host it with a robots.txt file telling Internet Archive not to archive it. Before the experiment I confirmed that Internet Archive had a history for this expired domain. After buying the domain and hosting the robots.txt file, I confirmed that Internet Archive no longer allowed access to it. I allowed the domain to expire, then went back to Internet Archive. The entire history of the domain was still there.

Experiment 2) I browsed the history of an existing website that I host to confirm they had it. Next I hosted a robots.txt file telling Internet Archive not to archive it and verified that the public can no longer browse the archive for this domain. Next I changed a picture on the website for six months, then changed it back. I waited another six months, removed the robots.txt file and checked the Internet Archive. I found that they had been taking snapshots throughout the year even though my robots.txt told them not to. The picture they were not supposed to have archived was visible in the archive.

If you really don't want them to archive your website, you can maybe block all of their IP addresses from accessing your server. Possibly by determining the domain name from the IP address and then checking if it is the Internet Archive.

Re: No

[Anonymous Coward]

Robots.txt is a suggestion, not a requirement.