Online directories that help people find advocates, read profiles and send enquiries are a form of legal technology, and like any online service they can fail quietly. A search page that loads slowly at the wrong moment, or a contact form that swallows messages, costs the visitor time and the platform its credibility. Site reliability engineering, or SRE, is the engineering discipline built to prevent that, yet it comes wrapped in myths. Six of them follow, taken one at a time, with the Ops Atlas entry on SRE as our guide. Nothing here is legal advice; it concerns how online services are run.

Myth 1: “SRE is just IT support with a fancier name”

It is a fair suspicion, because plenty of companies do exactly that: rename the help desk, keep the old routine and hope for the best. The Atlas has a name for it, “fake SRE”. The real discipline treats operations as a software problem. Engineers measure what visitors experience, set a numeric target, and write code that removes repetitive work. For a directory site, that might mean automatically checking that every city page returns results, rather than waiting for a visitor to complain.

Myth 2: “Our directory should aim for 100 percent uptime”

Sounds noble, but it is the wrong target for nearly everything. The Atlas explains that each further notch of dependability may cost around 100x the notch before, and visitors on a phone connection cannot tell whether a page was available 99.9 or 99.99 percent of the time. Better to choose a target such as 99.9 percent and spend the savings on better listings.

That choice has a concrete meaning. Over a 28-day window, 99.9 percent leaves a little over 40 minutes for things to go wrong, which equals a thousand bad requests in each million served. When something goes wrong, the platform has a known allowance to absorb it.

Myth 3: “More alerts mean more safety”

Quite the opposite. The Atlas lists alerting on every metric among the classic anti-patterns, because a team that gets pinged for everything soon ignores everything. A better method is to alert on how quickly the allowance is draining. A fast drain, such as a broken search page during peak hours, notifies a person at once. A slow drain, such as a few percent of slow responses overnight, becomes a ticket for the morning.

There is a subtle trap here. The entry adds that with a mere 90 percent objective, an alert keyed to a 2%-per-hour burn is mathematically unable to trigger, so objective and alert design must be settled together.

Myth 4: “Blameless reviews mean nobody is responsible”

Another favourite. A blameless postmortem looks for the conditions that allowed an error, rather than a person to punish. It is nevertheless strict about what happens next: every action has an owner and a date. The Atlas flags reviews whose actions never close as a failure pattern, and it suggests a trigger so that reviews are not optional. Whenever one incident swallows over twenty percent of the allowance, a written review must follow, carrying at least one urgent action.

For a directory, imagine a weekend when profile photos stopped loading across hundreds of pages. A good review would ask why no check noticed, what should alert next time, and who will build it by when.

Myth 5: “Once we have SRE, developers never have to carry a pager”

Not so. The entry observes that excess operational work is pushed back onto developers, so that the people who create a problem feel its cost. The Google guideline holds repetitive work below fifty percent of engineer hours, and a limit like that only bites if breaching it carries a cost, for instance moving developers back onto the rota. A smaller platform will not have a full rota, but the principle still holds: whoever can fix the cause should hear about it.

On sizing, the Atlas reckons about eight people are needed to cover one site day and night, with a cap of two incidents for each half-day shift. Few directories can staff that, so many borrow the ideas on a lighter scale, with a named person per week and clear escalation.

Myth 6: “SRE is only for giant technology companies”

Large companies popularised it, but the method scales down. The Atlas suggests a short screening before adoption: is there an outage visitors would genuinely feel, will someone senior put a signature on a policy, does manual work stay above half the team’s hours even after cuts, and can a sensible rota be staffed? If the answers are no, the entry’s logic is to wait, not to rename. A directory with a few thousand visitors a day can still define an indicator, a target and a written rule, using a spreadsheet.

What reliability looks like to a visitor

Visitors do not think in percentages. They judge by moments: the search that returned results instantly, the profile that opened, the enquiry form that confirmed receipt. That is why the discipline starts from the user’s seat and counts good events against valid ones. Compare the experience of reading a city listing, such as the page on immigration advocates in Madurai, with a family-law listing like the one for Coimbatore family advocates. A visitor expects both to open quickly and show consistent information, whatever the city. A reliable platform delivers that sameness, page after page.

Putting the myths to work

  1. Pick two journeys visitors care about, such as searching a city and sending an enquiry.
  2. Define what counts as a good attempt and count it for a month.
  3. Settle on an objective short of perfection and work out the allowance.
  4. Write a short rule for what pauses when the allowance runs low, and get it signed.
  5. Alert on burn rate, not on every metric.
  6. Review each bad day calmly, with owners and dates.

None of that involves buying anything. It does involve agreeing on numbers and keeping the record. As the Atlas puts it in effect, reliability is something to be spent deliberately rather than hoarded: a budget that stays nearly full can signal that the target is too loose or that useful improvements are being delayed.

Staff matter as much as software. A single named owner for each directory journey, even part-time, makes it far likelier that a quiet failure gets noticed. Add a monthly fifteen-minute look at the numbers and you have the skeleton of a practice.

A closing word

Each myth lingers because it is partly right. Renamed help desks do exist; perfect uptime is attractive; alerts do feel protective. The corrections are simply the discipline’s way of replacing feelings with evidence. If your platform serves people at stressful moments, evidence is worth having.

0 0 votes
Article Rating
Subscribe
Notify of
guest
0 Comments
Oldest
Newest Most Voted
0
Would love your thoughts, please comment.x
()
x