Blog

Dated commentary on what is happening in networking, security, and identity - written from a point of view, and true as of the day it was published.

12 pieces so far, posted as each is written, with no fixed cadence. What changes on the site itself is in the changelog.

  • Both ends of the escalation

    I have taken support cases and I have opened them: inside a vendor's support centre, as the vendor's escalation manager for two very large customers, and for a decade as the engineer opening cases on a partner's behalf. Two tools on this site come out of those years. This is the point of view they encode, stated in the first person.

  • Schrems III, seen from São Paulo: the triangle that decides where your data may live

    On 29 June the United States Supreme Court removed the independence of the agency that the EU-US data transfer framework cites 259 times. Five months earlier, Brazil and the European Union had recognised each other as adequate, the first such decision Brazil ever issued. The United States has none. For anyone in Brazil deploying a cloud security service, the same data now sits under three different legal positions depending on which way it flows - and the most common flow is the weakest.

  • SATAN to Mythos: the argument that comes back every time the tool gets better

    In 1995 a researcher was fired for publishing a program that found security holes. In 2026 a company withheld a model that finds them, gave it to twelve partners and forty organisations, and found nearly ten thousand severe flaws that mostly have not been fixed. The sentence in both arguments is the same. Almost everything else is not, and the differences are where a defender should be looking.

  • 264 hours: a first year of teaching, counted

    In 2020 the thing that had accompanied every job I held since 1997 became the job itself. The delivery was thirteen courses in thirty-three days, 264 hours, 51 students. What that number hides is the rest of the year, which went into being allowed to stand there at all, because in this profession you cannot teach a course you have not passed.

  • The only seat in the industry with no preference

    For most of a decade I supported resellers rather than customers, from inside distribution. It is the least visible job in the chain and it has one property nothing else in the industry has: a distributor carries competing vendors, so the engineer inside it is the only person who sees all of them without needing one to win.

  • One product, five flags: what happened to the box I used to demonstrate

    The appliance that invented clientless SSL VPN passed through five companies in twenty years. I worked with it at three of them: inside Juniper while Junos Pulse was taking shape, through the channel, and from the distribution side architecting it across Brazil. Under its fifth flag it belongs to the category the industry is now dismantling. That is not a story about a bad vendor. It is a story about a shape.

  • The certification that is not on the credentials page

    My oldest certification is a PADI Open Water card from April 1999, a year before my first CCNA. It is not on this site's credentials page and it never will be, which is worth a paragraph about what that page is for, and one about what diving and change control have in common.

  • The switch lineage, from inside a company that split into four

    I started in 1996 at Cabletron, one of networking's big four. In 2000 it broke into four companies, and I spent the decade after that working across the pieces: Riverstone, then Enterasys, then the equipment those companies made turning up under other logos. This is the switch's history told the way the founders article says histories are usually not told: from the seat of someone who stayed while the names changed around him.

  • Section 12 of a 1992 networking book: 'Network Management based on Artificial Intelligence'

    My first assignment as an intern in 1996 was to read a Cabletron guide to local area networking cover to cover. It was published in 1992. Its contents page lists nineteen sections, and the twelfth is about managing networks with artificial intelligence - rules-based against inductive modelling. I have the book. Thirty-four years later I spend my week teaching the products that argument eventually produced, and I am not sure the argument has moved as far as we think.

  • 1984, São Paulo: what Brazil's market reserve actually built

    For eight years Brazil legally reserved its computer market for domestic companies. The policy failed at what it said it was for. It also put a cloned Sinclair on my family's television in 1984, which is why I have the career I have. Writing as one of the things an industrial policy produced without meaning to.

  • The interviews that caught an industry in time

    Between 1988 and 1994 James L. Pelkey recorded eighty-one interviews with the people who built computer communications, while they were still working and still able to say why. The transcripts are public. A note on a piece of scholarship worth knowing about, and on what it is useful for.