Articles on Dan Shearer

Slow LLMs and MCPs are hiding problems

Thu, 12 Mar 2026 00:00:01 +0000

This is a technical note about a problem that is going to bite agentic AI users soon.

AI is slow, and Agentic AI is even slower. I develop an MCP server that generates PDF documents, and I work with the Agentic Perseverance Composition Engine daily, and AI seems so, so slow. There’s so much waiting, and every mistake means yet more sitting around. Tasks we know actually take maybe 5 microseconds on an operating system (eg, does a file called Things-to-Do exist?) can take one million time longer – between 2 and 5 seconds. This is because the big brain in the cloud is being consulted multiple times, often with timeouts. It’s a young, unstable and unreliable stack, rather like the early days of MS DOS or the Apple ][. When AI gets hold of the data from your computer via an MCP server it can do some very interesting things, but it is not put together well.

Data Mobility in the Trumpian Post-Brexit Era

Wed, 11 Feb 2026 05:55:38 +0000

From time to time I am engaged to help organisations in the UK and in the EU make decisions about where their data is stored, how it is accessed, and how to keep things as stable as possible over the next few years. This was a dizzying mess until 2025, and in 2026 there are some big decisions coming. Organisations need as much certainty as they can get for making decisions which are expensive to change in the future.

Code of Conduct

Mon, 09 Feb 2026 05:55:38 +0000

This file is a Code of Conduct first written in 2020 for the LumoSQL project. Here is Version 1.6 – Updated 9th February, 2026.

Heavily adapted and compressed from the large and repetitive version 3.1 of the Mozilla Participation Guidelines and published by LumoSQL under the Creative Commons Attribution-ShareAlike 4.0 International license.

LumoSQL Code of Conduct

This file exists because the LumoSQL Project needed it, less than one year after starting in 2019. We take it seriously, and hope that most English-reading adults can understand what is said. We hope this is not needed very often.

Opportunity in GDPR Article 28

Mon, 09 Feb 2026 05:55:38 +0000

The detail of the GDPR and its implied computer science contain a solution for sharing secrets according to law. This continues to be true in 2026, as the Digital Omnibus Regulation takes shape.

Executive Summary

The GDPR sets up a conflict in trust between companies in particular circumstances, which can only be resolved by using the automation of a cryptographic audit trail with particular properties as described below.

Problem Statement

Under the EU’s GDPR law virtually every company is a Controller, and virtually all Controllers use at least one Processor. When a Processor is engaged, the GDPR requires that a contract is signed with the very specific contents spelled out in clause 3 of Article 28. The GDPR requires that Controllers and Processors cooperate together in order to deliver data protection, and this cooperation needs to be very carefully managed to maintain the security and other guarantees that the GDPR also requires. That’s what this mandatory contract is intended to achieve.

Origins of EU-US privacy battles

Sun, 08 Feb 2026 07:00:00 +0000

Privacy Shield was a 2016 self-certification scheme for US companies to hold themselves to the strict EU privacy rules. In 2020 Privacy Shield was struck down by the EU Court of Justice. In non-technical terms, the Court said: There is no way Privacy Shield can work. So don’t use US-controlled cloud companies such as Google or Amazon.

In late 2021 this decision started rippling out across Europe, as one place and then another moves away from these giant US companies, starting with government users. We all like familiarity and wish to avoid change, so this decision seems astonishing to many people. Once organisations get over their surprise, it is not so difficult to do. In 2023 I wrote It remains to be seen what these US cloud companies will do in 2023. Some of them are wealthier than several smaller EU nations combined. By 2026 we had the answer - they used coercion by political, economic and legal means to prevent EU citizens using their own IP to build their own services.

Fossil

Sun, 08 Feb 2026 04:00:38 +0000

The Fossil source code management system is the most fully-featured alternative to Git, and has had twenty years of development and testing since 2006. After helping Fossil make some changes I now use Fossil for several projects. I also use Git extensively on various software forges (but not GitHub unless I must). Mercurial is actively maintained but has lost most of its mindshare since Mozilla, Bitbucket and others migrated away, and is rarely chosen for new projects today.

LumoSQL

Sat, 07 Feb 2026 05:55:38 +0000

LumoSQL protects data on mobile phones using a new data storage technology which is highly compatible with most existing devices. With LunmoSQL, the device owner has ultimate right to decide who can read or change their data… and this decision continues to be enforced even after it has been copied off the phone to (for example) a bank or insurance company for processing with their in-house database software. In contrast, the situation at present is that device owners are rarely in control of the privacy of their own data, despite many laws relating to privacy.

Patent process for Ballmer-era Microsoft Software Patents

Fri, 06 Feb 2026 11:55:39 +0000

I participated in many battles directly against Microsoft in the Ballmer era, 1998-2014. Every Samba feature release seem to further anger Microsoft. Copyright and then especially patents were weaponised, as well as well-funded hit teams aimed at spreading confusion and intimidating their own (Microsoft’s!) customers. In the Nadalla era from 2014-present, Microsoft and other tech giants are using even more brutal ways (paracopyright, technical protection measures and the Unitary Patent System), to coerce citizens and governments.

Copyright, patents, Samba and Microsoft

Fri, 06 Feb 2026 10:55:39 +0000

This timeline covers the period when Microsoft decided free software and Samba in particular was an exisential threat. Microsoft often buried competitors in expensive legislation, but turned out to be much more difficult to bury open source like Samba. This was the Ballmer era, named after the then-CEO, and the history of Samba’s triumphs feels highly relevant to 2026 where other giant companies seek to prevent the rise of open source competitors.

In 2014, Microsoft got a new CEO and dramatically changed course from explicit hostility to embracing open source. The battleground is now about paracopyright and preventing non-US cloud but it has its roots in the great open source IP battles of the 21st century.

Samba

Wed, 04 Feb 2026 13:55:38 +0000

In 2026, the Samba Project is nearly 30 years old and has conservatively a billion users. Samba started when I got upset at Microsoft for trying to monopolise all computer networking. I discovered some unmaintained but interesting open source software for sharing files and printers with workstation computers. And the rest is the official Samba history.

Samba is implemented by talented software engineers with a very large number of total contributors. I was (and remain) most interested in interoperability architecture and design, why these things are needed and make sense to users. Plus some protocol analysis, for example, technical readers may know the NTLMv2 encryption scheme was tricky, but turned out to be the same as used in the NTFS filesystem - NTLM is deprecated in favour of Kerberos now but those were the days. I wrote How to Replace Windows NT with Linux, explaining protocol-first strategies for removing Microsoft software.

BibLaTeX, eras and scripts

Tue, 03 Feb 2026 17:55:38 +0000

How to manage BibLaTeX across time and cultures

I wrote a paper in English using LaTeX on the topic of Epidemiology and One Health. Some essential references did not exist in English. That might sound simple — just list the originals, plus some translation/cross-referencing work to get the necessary information! It isn’t that simple.

This howto is for LaTeX authors with references which are less common in computing/mathematics but otherwise unremarkable, particularly: non-latin scripts, latinisations, non-English references, rare scripts and ancient documents. My sources had all of these at once, giving me the following situation:

Logical and Thermodynamic Reversibility

Sun, 01 Feb 2026 05:55:38 +0000

Large Language Models are subject to the laws of physics in a bad way, because they use so much electricity and make so much heat. I was interested to learn about a Mr Landauer and his principle of thermodynamic reversibility, which suggests physics just might come to our rescue and greatly reduce the amount of power required by AI datacentres. (Note ’logical reversibility’ sounds confusingly like reversible computers and backwards execution, but apart from a general spirit of going backwards, they are completely unrelated.)

Not Forking

Sun, 01 Feb 2026 05:55:38 +0000

Not-forking is a technical tool for software development. Not-forking assists with reproducibility.

Here are some simple ways of explaining what Not-forking can do:

Not-forking lets you integrate non-diffable codebases, a bit like patch/sed/diff/cp/mv rolled into one.
Not-forking is a machine-readable file format and tool. It answers the question: What is the minimum difference between multiple source trees, and how can this difference be applied as versions change over time?
Not-forking avoids duplicating source code. When one project is within another project, and the projects are external to each other, there is often pressure to fork the inner project. Not-forking avoids that.
Not-forking helps address the problem of reproducibility. By giving much better control over the input source trees, it is more likely that the output binaries are the same each time.

But here is the big win: Not-forking avoids project-level forking by largely automating change management in ways that version control systems such as Fossil, Git, or GitHub cannot.
The full documentation goes into much more detail than this overview.

Reversible Execution

Sun, 01 Feb 2026 05:55:38 +0000

Reversible execution creates computers that seem to run backwards, applying time shifting techniques with simulation/virtualisation. They address the problems of software unreliability and complexity, and I believe my excited comments from 2005 still stand:

Reversibility is the biggest advance in debugging since source code debugging

— Me, on the GDB developers list

In 2026, reversibility still isn’t seen as an ubiquitous must-have for software development, but awareness is increasing.

At the same time, the equally interesting topics of logical reversibility and thermodynamic reversibility have become very important. They are not really anything to do with reversible execution, they just sound similar. But if you’re interested in the problems AI datacentres present the world, this kind of reversibility looks highly relevant.

Security Standards and Certifications

Sun, 01 Feb 2026 05:55:38 +0000

I have been lead implementer of the main security and privacy standards several times each. These can seem intimidating, but properly used they improve security overall, and can help a business run more smoothly.

From a pragmatic, business point of view:

These standards are about writing down the actual rules of your business relevant to security and privacy, and then writing down how you improve these rules, and recording how well they work. All businesses can benefit from challenging their working habits and practices, and since privacy and security touch most parts of a business, this is an opportunity to review how the business works before something goes wrong. From the point of view of both Computer Science and Information Management Science:

Education Exercises

Sun, 01 Feb 2026 02:55:38 +0000

These are some exercises and tricks I have either created or been subjected to over the years, and I have mentored students through them on many occasions.

The general theme here is that most of the systems and stacks that are taken for granted often don’t work very well, and often don’t seem to have a very bright future. This is even the case for famous codebases relied on by billions of people. There are no absolutes and no immediate fixes, but it is food for thought if we can demonstrate immense waste of human effort amid poor quality computing systems, even when impressive modern computer science is applied.

Articles on Dan Shearer

Slow LLMs and MCPs are hiding problems

Data Mobility in the Trumpian Post-Brexit Era

Code of Conduct

Contents

LumoSQL Code of Conduct

Opportunity in GDPR Article 28

Origins of EU-US privacy battles

Fossil

LumoSQL

Patent process for Ballmer-era Microsoft Software Patents

Copyright, patents, Samba and Microsoft

Samba

BibLaTeX, eras and scripts

How to manage BibLaTeX across time and cultures

Logical and Thermodynamic Reversibility

Not Forking

Reversible Execution

Security Standards and Certifications

Education Exercises