Guide
Podcast: should audio hosting live with the website?
Putting 80 MB MP3 files on the same shared host as WordPress saturates bandwidth and backups. RSS, audio CDN and dedicated platforms — a different calculation from a brochure site.
No results for this search
Guide
cPanel, Plesk, or none: which control panel fits your autonomy?
cPanel reassures beginners; Plesk unifies Windows and Linux; without a panel you gain control and lose the safety net. The right choice follows your technical autonomy — not fashion.
Guide
Giving SFTP access to a team without sharing one password
One shared FTP account in LastPass means zero traceability and impossible revocation. Individual accounts, SSH keys and folder-level rights — minimum hygiene before the first freelancer.
Guide
Domain redirects: preserve SEO and journeys without creating loops
HTTP 301, registrar change, www vs apex — a bad redirect erases SEO or traps users in an infinite loop. Map before touching DNS.
Guide
Mailing lists: why your web host is not always the right sender
Sending 10,000 newsletters from the site shared SMTP invites spam folders and suspension. Deliverability, IP reputation, and compliance are a different job than web hosting.
Guide
IPv6 for a new website: enable it without making it a separate project
Dual-stack A + AAAA is enough for most new sites. No need to rebuild your infrastructure — enable IPv6 at the host, check DNS, and test mail separately.
Guide
Choosing a cloud region: latency, law, and cost in one calculation
Gravelines looks close and cheap — until Swiss users wait 80 ms more and backups replicate to the US without a data processing agreement. A region is chosen in three dimensions, never at random.
Guide
Disaster recovery: turning backups into real continuity
Having a backup is not a recovery plan. RTO, RPO, tested restore, and defined roles — what separates a disk copy from an organisation that survives a major outage.
Guide
Abuse reports: prepare a response before the first message
Phishing, spam, flagged content — the host opens a ticket with a twenty-four-hour deadline. Without an internal procedure, you lose the site before finding the compromised account.
Guide
Host, client, agency: where does each responsibility begin?
Outage, data breach or illegal content — everyone points elsewhere. Without a written matrix, the client assumes "everything is included" and the host points to "your application".
Guide
Hosting terms: clauses to read before you need them
"Best effort" backups, suspension without notice, data ownership — hosting terms are read on signing day, not on unpaid invoice or dispute day.
Guide
Static site: can the simplest choice last?
Hugo, Eleventy or plain HTML — little server, small attack surface. Until you need auth, search or a thousand pages per day. Where static holds — and where it breaks.
Guide
Jamstack: publishing fast without forgetting dynamic content
Static builds plus a CDN speed delivery — but search, accounts, checkout, and editorial preview stay dynamic. Map those flows before celebrating the Git deploy.
Guide
Headless CMS: when decoupling the front end truly frees the project
Strapi, Directus or Contentful separate content and rendering — useful with multiple fronts and autonomous front team. Otherwise you pay for two stacks for a brochure site.
Guide
Jupyter for a team: making notebooks shareable and safe
A notebook running locally on 32 GB RAM is not team infrastructure. Sharing, auth, sensitive data and GPU cost — how to industrialise without Jupyter becoming shadow IT.
Guide
Renting a GPU for AI: questions that avoid an absurd bill
Hourly or monthly, spot or reserved, real or shared VRAM — a badly chosen GPU instance costs more idle than used. Ask these questions before signing.
Guide
Hosting a language model: start with constraints, not GPUs
Renting a GPU before defining latency, context, and confidentiality leads to GPU bills and a POC that never reaches production. Map usage, data, and budget first.
Guide
Communicating during an outage: say what you know without guessing
A vague message reassures nobody; a false ETA destroys trust. In an incident, structure updates — impact, scope, actions, next checkpoint — without guessing the cause.
Guide
A useful status page before an incident, not only after
A status page built in five minutes after an outage helps nobody. Prepare components, subscribers and runbook before the first 502 — your customers and support will thank you.
Guide
The 99.99% myth: translating a percentage into lost minutes
Four nines means 52 minutes of downtime per year — in theory. In practice, SLAs exclude maintenance, DNS, and sometimes upstream network. Read the contract, not the badge.
Guide
Core Web Vitals: what a host can improve—and what it cannot
LCP, INP and CLS depend as much on your theme as on the server. Before switching hosts for a PageSpeed badge, separate what belongs to CDN, PHP and the front end.
Guide
HTTP/3: visible speedup or premature optimization?
HTTP/3 and QUIC promise lower latency on unstable networks. On a site already fast over HTTP/2 with healthy TTFB, gains are often marginal — sometimes negative if origin stays slow.
Guide
OPcache: the PHP setting worth checking first
OPcache stops recompiling PHP on every request — immediate CPU and latency gain. Misconfigured or disabled after deploy, it stays invisible until the next outage.
Guide
PHP-FPM: tuning workers without turning the server into a gamble
Too few PHP-FPM workers and requests queue; too many and RAM collapses. Here is how to calculate pm.max_children from real memory use, not a magic rule.