Dev.to Security πŸ” Cybersecurity πŸ‘ 0 πŸ“– 5 min read

52 outbound links, three rel policies nobody wrote down, and the only four we let pass link equity are our own apps

Most of notifio.app is pages about other people's websites. /alerts/rightmove describes what Rightmove's own property alerts do; /compare/rentbird puts our one-off price next to a subscription. Pages like that make factu

Most of notifio.app is pages about other people's websites. /alerts/rightmove describes what Rightmove's own property alerts do; /compare/rentbird puts our one-off price next to a subscription. Pages like that make factual claims about companies that never asked to be written about, so every claim carries a source link rendered on the page.

I have counted those links twice. 45 citations on our site, 34 distinct URLs counted them in the source. I curled the 49 outbound links on our site checked whether they still answered. Both times I treated a link as a URL. I never read the attributes around it, which is where a site's actual editorial policy lives.

The measurement

38 pages fetched from the live site, every anchor parsed, internal links discarded:

for m in re.finditer(r'<a\b([^>]*)>', html):
    href = re.search(r'href="([^"]*)"', m.group(1))
    if not href or not href.group(1).startswith("http"): continue
    if "notifio.app" in href.group(1): continue
    rel = re.search(r'\brel="([^"]*)"', m.group(1))
    tgt = re.search(r'\btarget="([^"]*)"', m.group(1))
  • 52 external anchor instances
  • on 28 of 38 pages, so ten pages link nowhere off-site at all
  • 40 distinct URLs across 32 distinct domains
  • 3 distinct rel values

Three. On a site where I would have said there was one.

rel target Count What they are
nofollow noopener _blank 45 citations on the site, comparison and guide pages
noopener noreferrer _blank 3 Stripe and Resend privacy policies, in our own /privacy
none none 4 our four sibling apps, on /about

The 45 break down as 25 on the fifteen /alerts pages, 6 on the five /compare pages and 14 across the /guides and /for articles, which matches the source census exactly.

Where three policies came from

$ grep -rn 'rel="' src/
src/app/privacy/page.tsx:171:          rel="noopener noreferrer"
src/app/privacy/page.tsx:191:          rel="noopener noreferrer"
src/app/privacy/page.tsx:206:          rel="noopener noreferrer"
src/app/alerts/[site]/page.tsx:135:   rel="nofollow noopener"
src/app/compare/[target]/page.tsx:176: rel="nofollow noopener"
src/components/ArticleLayout.tsx:154:  rel="nofollow noopener"

Six occurrences in four files. The 45 come from three template files that each independently spell out the same string, the 3 are hand-typed one per paragraph in a legal page, and the 4 on /about are a hand-written anchor with no rel at all:

<a href="https://cogniprep.app" class="underline underline-offset-4 hover:text-[#fafafa]">

There is a smaller tell in there too. The citation templates write rel before target; the privacy page writes target before rel. Same author, different months.

Read back, the policy is coherent

Nobody wrote this down, and it still says something consistent:

  • A competitor gets a citation and no endorsement. nofollow, new tab.
  • A subprocessor gets no referrer. noreferrer means Stripe's logs do not record which page of ours sent a reader to their privacy policy.
  • Our own apps get everything. Followed, same tab, treated as part of one estate rather than as the outside world. That is the only group whose links can pass ranking signal, and I own all four destinations.

I will take that. If I had sat down to write an outbound link policy for a small product site, I would have written approximately that.

Then there are the two omissions, which mirror each other

The 45 citation links are nofollow with no noreferrer. So every rival we write about can see, in their own analytics, exactly which page of ours sent that visitor. They get the referrer and none of the ranking value.

The 3 processor links are noreferrer with no nofollow. So Stripe and Resend get the ranking value and none of the referrer.

Each group has precisely what the other is missing, and in both cases the thing that was left out is the thing the destination would have wanted. Neither of those was a decision. They are two different afternoons.

The security half of this is fine, almost by accident. 48 of the 52 open a new tab and all 48 carry noopener, so there is no reverse tabnabbing surface. The 4 without noopener are the same 4 without target="_blank", so they never hand a window reference to anybody either. Belt and braces, given that current browsers imply noopener for target="_blank" anyway.

The question I now have to answer

The uncomfortable one is whether those 45 should be nofollow at all.

Google's own framing is that nofollow is for links you cannot vouch for and sponsored is for paid placement. A citation that exists specifically so a reader can verify a factual claim about somebody's product is close to the opposite of that. It is the most sincere kind of link on the web, and the reason these pages have any credibility is that the claims are checkable. Marking all 45 nofollow means taking the credibility of citing a source while routing none of the value to the source.

There is a real argument the other way for the six on /compare, where the destination is a rival's commercial pricing page and the page it sits on exists to win the comparison. Nofollowing that is defensible.

It is much harder to defend for the 25 on /alerts, which mostly point at help centre articles explaining how that site's own alerts work: Rightmove's own documentation of its alert dispatch, Kamernet's "how does it work", SpareRoom's explanation of its early bird feature. Those are the documents the page is reasoning about. A link to a company's own help page is not an endorsement of anything except the fact that they wrote it.

So the thing I am actually going to do is the boring part first: those three identical rel strings in three templates should be one exported constant, because a policy that lives in three files is a policy that will drift the next time I add a page type. Then the question of what the constant should contain is one decision in one place, instead of a string I have to remember to change three times while telling myself I have a policy.

The pages all this is about are at notifio.app/alerts and notifio.app/compare, the four followed links are at the bottom of notifio.app/about, and the app itself is at notifio.app/download.

πŸ“° Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β€” full credit and traffic to the original publisher.