Is PageRank Lost on Links to Binary Files (e.g. ZIP, EXE, TAR)?

[robzilla]

The recent commotion surrounding Google's change of policy regarding nofollow got me thinking about what other types of links PageRank may be lost on. I ended up with a question that I've previously asked myself: do pages that include direct links to binary files, such as .exe and .zip, lose a portion of their PageRank to those links, or does the fact (read: my assumption) that these files do not exist on the link graph mean that the links won't be entered into the PageRank calculation?

[rainborick]

Since the file types you mention are not indexed, they wouldn't actually receive any PageRank, but I suppose that if you linked to them with an <a>nchor tag the PageRank that would ordinarily be sent by such a link would be "wasted". I don't think there's been any official comment on this situation, so it's just a guess. There's always a chance that Google reassigns the PageRank when it eventually determines that a link points to a non-indexable file.

If you're truly concerned about it, you can always use a <form> tag for such files to allow users to access them without any loss of PageRank, but I wouldn't put a lot of effort into it.

[Lord Majestic]

Since the file types you mention are not indexed, they wouldn't actually receive any PageRank

Such URLs called "dangling links" in PageRank algorithm.

They don't need to be indexed in order to have PageRank, however they DO need to have outgoing urls (and thus indexed) if they were to pass PageRank further.

[steveb]

Link to http://www.example.com/test.exe

301 redirect that URL to http://www.example.com/test.html

See the html page get pagerank.

[Lord Majestic]

Redirect can be viewed as having a page with just one outgoing link, so in theory all PageRank should flow to the target page (in practice permanent redirect might be needed).

[tedster]

in theory all PageRank should flow to the target page

Even the originally published PR formula includes a damping factor (0.85 was suggested in the paper). Otherwise, as the calculation is iterated, the PR values would take off to infinity rather than converge.

The key is that a PR "vote" is not actually defined as a one-time event, "slice up the available PR and add it to the target page". PR is defined as an iterative calculation -- something repeated around the webgraph many times until value converge. So even theoretically, "all" the PageRank never flows through a page or a redirect.

[Lord Majestic]

Even the originally published PR formula includes a damping factor (0.85 was suggested in the paper). Otherwise, as the calculation is iterated, the PR values would take off to infinity rather than converge.

Indeed - by saying "all", I did not intend to mean all PageRank page gets but all that is being flowed down the line - 0.85 in case of classic PageRank, which is much more than 0 which was asserted above.

[robzilla]

There's always a chance that Google reassigns the PageRank when it eventually determines that a link points to a non-indexable file.

Would it make sense for Google to do that? To reassign the PageRank, instead of throwing it out?

[steveb]

When talking about gazabillions of links, and the fact the destinations might be indexable again a day later, it doesn't seem like the most sensible way to do it. Looking at all deadends as simply deadends is consistent.

I suppose though they have tested it several ways, and maybe still do test it.