TOPIC: COMPUTING
Mixing local and cloud capabilities in an AI toolkit
9th September 2025
The landscape of AI development is shifting towards systems that prioritise local control, privacy and efficient resource management whilst maintaining the flexibility to integrate with external services when needed. This guide explores how to build a comprehensive AI toolkit that balances these concerns through seven key principles: local-first architecture, privacy preservation, standardised tool integration, workflow automation, autonomous agent development, efficient resource management and multi-modal knowledge handling.
Local-First Architecture and Control
The foundation of a robust AI toolkit begins with maintaining direct control over core components. Rather than relying entirely on cloud services, a local-first approach provides predictable costs, enhanced privacy and improved reliability whilst still allowing selective use of external resources.
Llama-Swap exemplifies this philosophy as a lightweight proxy that manages multiple language models on a single machine. This tool listens for OpenAI-style API calls, inspects the model field in each request, and ensures that the correct backend handles that call. The proxy intelligently starts or stops local LLM servers so only the required model runs at any given time, making efficient use of limited hardware resources.
Setting up this system requires minimal infrastructure: Python 3, Homebrew on macOS for package management, llama.cpp for hosting GGUF models locally and the Hugging Face CLI for model downloads. The proxy itself is a single binary that can be configured through a simple YAML file, specifying model paths and commands. This approach transforms model switching from a manual process of stopping and starting different servers into a seamless experience where clients can request different models through a single port.
The local-first principle extends beyond model hosting. Obsidian demonstrates this with its markdown-based knowledge management system that stores everything locally whilst providing rich linking capabilities and plugin extensibility. This gives users complete control over their data, whilst maintaining the ability to sync across devices when desired.
Privacy and Data Sovereignty
Privacy considerations permeate every aspect of AI toolkit design. Local processing inherently reduces exposure of sensitive data to external services, but even when cloud services are necessary, careful evaluation of data handling practices becomes crucial.
Voice processing illustrates these concerns clearly. ElevenLabs offers high-quality text-to-speech and voice cloning capabilities but requires careful assessment of consent and security policies when handling voice data. Similarly, services like NoteGPT that process documents and videos must be evaluated against regional regulations such as GDPR, particularly when handling sensitive information.
The principle of data minimisation suggests using local processing wherever feasible and cloud services only when their capabilities significantly outweigh privacy concerns. This might mean running smaller language models locally for routine tasks, whilst reserving larger cloud models for complex reasoning that exceeds local capacity.
Tool Integration and Standardisation
As AI systems become more sophisticated, the ability to integrate diverse tools through standardised protocols becomes essential. The Model Context Protocol (MCP) addresses this need by defining how lightweight servers present databases, file systems and web services to AI models in a secure, auditable manner.
MCP servers act as bridges between AI models and real systems, whilst MCP clients are applications that discover and utilise these servers. This standardisation enables a rich ecosystem of tools that can be mixed and matched according to specific needs.
Several clients demonstrate different approaches to MCP integration. Claude Desktop auto-starts configured servers on launch, making tools immediately available. Cursor AI integrates MCP servers directly into coding environments, allowing function calls to route to custom servers automatically. Continue provides open-source alternatives for VS Code and JetBrains, whilst LibreChat offers a flexible chat interface that can connect to various model providers and MCP servers.
The standardisation extends to development workflows through tools like Claude Code, which integrates with GitHub repositories to automate routine tasks. By creating a Claude GitHub App, developers can use natural language comments to trigger actions like generating Docker configurations, reviewing code or updating documentation.
Workflow Automation and Productivity
Effective AI toolkits streamline repetitive tasks and augment human decision-making, rather than replacing it entirely. This automation spans from simple content generation to complex research workflows that combine multiple tools and services.
A practical research workflow demonstrates this integration. Beginning with a focused question, Perplexity AI can generate citation-backed reports using its deep research capability. These reports, exported as PDFs, can then be uploaded to NotebookLM for interactive exploration. NotebookLM transforms static content into searchable material, generates audio overviews that render complex topics as podcast-style conversations, and builds mind maps to reveal relationships between concepts.
This multi-stage process turns surface reading into grounded understanding by enabling different modes of engagement with the same material. The automation handles the mechanical aspects of research synthesis, whilst preserving human judgement about relevance and interpretation.
Repository management represents another automation frontier. GitHub integrations can handle issue triage, code review, documentation updates and refactoring through natural language instructions. This reduces cognitive overhead for routine maintenance whilst maintaining developer control over significant decisions.
Agentic AI and Autonomous Systems
The evolution from reactive prompt-response systems to goal-oriented agents represents a fundamental shift in AI system design. Agentic systems can plan across multiple steps, initiate actions when conditions warrant, and pursue long-running objectives with minimal supervision.
These systems typically combine several architectural components: a reasoning engine (usually an LLM with structured prompting), memory layers for preserving context, knowledge bases accessible through vector search and tool interfaces that standardise how agents discover and use external capabilities.
Patterns like ReAct interleave reasoning steps with tool calls, creating observe-think-act loops that enable continuous adaptation. Modern AI systems employ planning-first agents that formulate strategies before execution and adapt dynamically, alongside multi-agent architectures that coordinate specialist roles through hierarchical or peer-to-peer protocols.
Practical applications illustrate these concepts clearly. An autonomous research agent might formulate queries, rank sources, synthesise material and draft reports, demonstrating how complex goals can be decomposed into manageable subtasks. A personal productivity assistant could manage calendars, emails and tasks, showing how agents can integrate with external APIs whilst learning user preferences.
Safety and alignment remain paramount concerns. Constraints, approval gates and override mechanisms guard against harmful behaviour, whilst feedback mechanisms help maintain alignment with human intent. The goal is augmentation rather than replacement, with human oversight remaining essential for significant decisions.
Resource Management and Efficiency
Efficient resource utilisation becomes critical when running multiple AI models and services on limited hardware. This involves both technical optimisation and strategic choices about when to use local versus cloud resources.
Llama-Swap's selective concurrency feature exemplifies intelligent resource management. Whilst the default behaviour runs only one model at a time to conserve resources, groups can be configured to allow several smaller models to remain active together whilst maintaining swapping for larger models. This provides predictable resource usage without sacrificing functionality.
Model quantisation represents another efficiency strategy. GGUF variants of models like SmolLM2-135M-Instruct and Qwen2.5-0.5B-Instruct can run effectively on modest hardware whilst still providing distinct capabilities for different tasks. The trade-off between model size and capability can be optimised for specific use cases.
Cloud services complement local resources by handling computationally intensive tasks that exceed local capacity. The key is making these transitions seamless, so users can benefit from both approaches without managing complexity manually.
Multi-Modal Knowledge Management
Modern AI toolkits must handle diverse content types and enable fluid transitions between different modes of interaction. These span text processing, audio generation, visual content analysis and format conversion.
NotebookLM demonstrates sophisticated multi-modal capabilities by accepting various input formats (PDFs, images, tables) and generating different output modes (summaries, audio overviews, mind maps, study guides). This flexibility enables users to engage with information in ways that match their learning preferences and situational constraints.
NoteGPT extends this concept to video and presentation processing, extracting transcripts, segmenting content and producing summaries with translation capabilities. The challenge lies in preserving nuance during automated processing whilst making content more accessible.
Integration between different knowledge management approaches creates additional value. Notion's workspace approach combines notes, tasks, wikis and databases with recent additions like email integration and calendar synchronisation. Evernote focuses on mixed media capture and web clipping with cross-platform synchronisation.
The goal is creating systems that can capture information in its natural format, process it intelligently, and present it in ways that facilitate understanding and action.
Conclusion
Building an effective AI toolkit requires balancing multiple concerns: maintaining control over sensitive data whilst leveraging powerful cloud services, automating routine tasks whilst preserving human judgement, and optimising resource usage whilst maintaining system flexibility. The market demand for these skills is growing rapidly, with companies actively seeking professionals who can implement RAG systems, build reliable agents and manage hybrid AI architectures.
The local-first approach provides a foundation for this balance, giving users control over their data and computational resources whilst enabling selective integration with external services. RAG has evolved from a technical necessity for small context windows to a strategic choice for cost reduction and reliability improvement. Standardised protocols like MCP make it practical to combine diverse tools without vendor lock-in. Workflow automation reduces cognitive overhead for routine tasks, and agentic capabilities enable more sophisticated goal-oriented behaviour.
Success depends on thoughtful integration rather than simply accumulating tools. The most effective systems combine local processing for privacy-sensitive tasks, cloud services for capabilities that exceed local resources, and standardised interfaces that enable experimentation and adaptation as needs evolve. Whether the goal is reducing API costs through efficient RAG implementation or building agents that prevent hallucinations through grounded retrieval, the principles remain consistent: maintain control, optimise resources and preserve human oversight.
This approach creates AI toolkits that are not only adaptable, secure and efficient but also commercially viable and career-relevant in a rapidly evolving landscape where the ability to build reliable, cost-effective AI systems has become a competitive necessity.
Advance your Data Science, AI and Computer Science skills using these online learning opportunities
25th July 2025
The landscape of online education has transformed dramatically over the past decade, creating unprecedented access to high-quality learning resources across multiple disciplines. This comprehensive examination explores the diverse array of courses available for aspiring data scientists, analysts, and computer science professionals, spanning from foundational programming concepts to cutting-edge artificial intelligence applications.
Data Analysis with R Programming
R programming has established itself as a cornerstone language for statistical analysis and data visualisation, making it an essential skill for modern data professionals. DataCamp's Data Analyst with R programme represents a comprehensive 77-hour journey through the fundamentals of data analysis, encompassing 21 distinct courses that progressively build expertise. Students begin with core programming concepts including data structures, conditional statements, and loops before advancing to sophisticated data manipulation techniques using tools such as dplyr and ggplot2. The curriculum extends beyond basic programming to include R Markdown for reproducible research, data manipulation with data.table, and essential database skills through SQL integration.
For those seeking more advanced statistical expertise, DataCamp's Statistician with R career track provides an extensive 108-hour programme spanning 27 courses. This comprehensive pathway develops essential skills for professional statistician roles, progressing from fundamental concepts of data collection and analysis to advanced statistical methodology. Students explore random variables, distributions, and conditioning through practical examples before advancing to linear and logistic regression techniques. The curriculum encompasses sophisticated topics including binomial and Poisson regression models, sampling methodologies, hypothesis testing, experimental design, and A/B testing frameworks. Advanced modules cover missing data handling, survey design principles, survival analysis, Bayesian data analysis, and factor analysis, making this track particularly suitable for those with existing R programming knowledge who seek to specialise in statistical practice.
The Google Data Analytics Professional Certificate programme, developed by Google and hosted on Coursera with US and UK versions, offers a structured six-month pathway for those seeking industry-recognised credentials. Students progress through eight carefully designed courses, beginning with foundational concepts in "Foundations: Data, Data, Everywhere" and culminating in a practical capstone project. The curriculum emphasises real-world applications, teaching students to formulate data-driven questions, prepare datasets for analysis, and communicate findings effectively to stakeholders.
Udacity's Data Analysis with R course presents a unique proposition as a completely free resource spanning two months of study. This programme focuses intensively on exploratory data analysis techniques, providing students with hands-on experience using RStudio and essential R packages. The course structure emphasises practical application through projects, including an in-depth exploration of diamond pricing data that demonstrates predictive modelling techniques.
Advanced Statistical Learning and Specialised Applications
Duke University's Statistics with R Specialisation elevates statistical understanding through a comprehensive seven-month programme that has earned a 4.6-star rating from participants. This five-course sequence delves deep into statistical theory and application, beginning with probability and data fundamentals before progressing through inferential statistics, linear regression, and Bayesian analysis. The programme distinguishes itself by emphasising both theoretical understanding and practical implementation, making it particularly valuable for those seeking to master statistical concepts rather than merely apply them.
The R Programming: Advanced Analytics course on Udemy, led by instructor Kirill, provides focused training in advanced R techniques within a compact six-hour format. This course addresses specific challenges that working analysts face, including data preparation workflows, handling missing data through median imputation, and working with complex date-time formats. The curriculum emphasises efficiency techniques such as using apply functions instead of traditional loops, making it particularly valuable for professionals seeking to optimise their analytical workflows.
Complementing this practical approach, the Applied Statistical Modelling for Data Analysis in R course on Udemy offers a more comprehensive 9.5-hour exploration of statistical methodology. The curriculum covers linear modelling implementation, advanced regression analysis techniques, and multivariate analysis methods. With its emphasis on statistical theory and application, this course serves those who already possess foundational R and RStudio knowledge but seek to deepen their understanding of statistical modelling approaches.
Imperial College London's Statistical Analysis with R for Public Health Specialisation brings academic rigour to practical health applications through a four-month programme. This specialisation addresses real-world public health challenges, using datasets that examine fruit and vegetable consumption patterns, diabetes risk factors, and cardiac outcomes. Students develop expertise in linear and logistic regression while gaining exposure to survival analysis techniques, making this programme particularly relevant for those interested in healthcare analytics.
Visualisation and Data Communication
Johns Hopkins University's Data Visualisation & Dashboarding with R Specialisation represents the pinnacle of visual analytics education, achieving an exceptional 4.9-star rating across its four-month curriculum. This five-course programme begins with fundamental visualisation principles before progressing through advanced ggplot2 techniques and interactive dashboard development. Students learn to create compelling visual narratives using Shiny applications and flexdashboard frameworks, skills that are increasingly essential in today's data-driven business environment.
The programme's emphasis on publication-ready visualisations and interactive dashboards addresses the growing demand for data professionals who can not only analyse data but also communicate insights effectively to diverse audiences. The curriculum balances technical skill development with design principles, ensuring graduates can create both statistically accurate and visually compelling presentations.
Professional Certification Pathways
DataCamp's certification programmes offer accelerated pathways to professional recognition, with each certification designed to be completed within 30 days. The Data Analyst Certification combines timed examinations with practical assessments to evaluate real-world competency. Candidates must demonstrate proficiency in data extraction, quality assessment, cleaning procedures, and metric calculation, reflecting the core responsibilities of working data analysts.
The Data Scientist Certification expands these requirements to include machine learning and artificial intelligence applications, requiring candidates to collect and interpret large datasets whilst effectively communicating results to business stakeholders. Similarly, the Data Engineer Certification focuses on data infrastructure and preprocessing capabilities, essential skills as organisations increasingly rely on automated data pipelines and real-time analytics.
The SQL Associate Certification addresses the universal need for database querying skills across all data roles. This certification validates both theoretical knowledge through timed examinations and practical application through hands-on database challenges, ensuring graduates can confidently extract and manipulate data from various database systems.
Emerging Technologies and Artificial Intelligence
The rapid advancement of artificial intelligence has created new educational opportunities that bridge traditional data science with cutting-edge generative technologies. DataCamp's Understanding Artificial Intelligence course provides a foundation for those new to AI concepts, requiring no programming background whilst covering machine learning, deep learning, and generative model fundamentals. This accessibility makes it valuable for business professionals seeking to understand AI's implications without becoming technical practitioners.
The Generative AI Concepts course builds upon this foundation to explore the specific technologies driving current AI innovation. Students examine how large language models function, consider ethical implications of AI deployment, and learn to maximise the effectiveness of AI tools in professional contexts. This programme addresses the growing need for AI literacy across various industries and roles.
DataCamp's Large Language Model Concepts course provides intermediate-level exploration of the technologies underlying systems like ChatGPT. The curriculum covers natural language processing fundamentals, fine-tuning techniques, and various learning approaches including zero-shot and few-shot learning. This technical depth makes it particularly valuable for professionals seeking to implement or customise language models within their organisations.
The ChatGPT Prompt Engineering for Developers course addresses the developing field of prompt engineering, a skill that has gained significant commercial value. Students learn to craft effective prompts that consistently produce desired outputs from language models, a capability that combines technical understanding with creative problem-solving. This expertise has become increasingly valuable as organisations integrate AI tools into their workflows.
Working with OpenAI API provides practical implementation skills for those seeking to build AI-powered applications. The course covers text generation, sentiment analysis, and chatbot development, giving students hands-on experience with the tools that are reshaping how businesses interact with customers and process information.
Computer Science Foundations
Stanford University's Computer Science 101 offers an accessible introduction to computing concepts without requiring prior programming experience. This course addresses fundamental questions about computational capabilities and limitations whilst exploring hardware architecture, software development, and internet infrastructure. The curriculum includes essential topics such as computer security, making it valuable for anyone seeking to understand the digital systems that underpin modern society.
The University of Leeds' Introduction to Logic for Computer Science provides focused training in logical reasoning, a skill that underlies algorithm design and problem-solving approaches. This compact course covers propositional logic and logical modelling techniques that form the foundation for more advanced computer science concepts.
Harvard's CS50 course, taught by Professor David Malan, has gained worldwide recognition for its engaging approach to computer science education. The programme combines theoretical concepts with practical projects, teaching algorithmic thinking alongside multiple programming languages including Python, SQL, HTML, CSS, and JavaScript. This breadth of coverage makes it particularly valuable for those seeking a comprehensive introduction to software development.
MIT's The Missing Semester course addresses practical tools that traditional computer science curricula often overlook. Students learn command-line environments, version control with Git, debugging techniques, and security practices. These skills prove essential for professional software development but are rarely taught systematically in traditional academic settings.
Accessible Learning Resources and Community Support
The democratisation of education extends beyond formal courses to include diverse learning resources that support different learning styles and schedules. YouTube channels such as Programming with Mosh, freeCodeCamp, Alex the Analyst, Tina Huang, and Ken Lee provide free, high-quality content that complements formal education programmes. These resources offer everything from comprehensive programming tutorials to career guidance and project-based learning opportunities.
The 365 Data Science platform contributes to this ecosystem through flashcard decks that reinforce learning of essential terminology and concepts across Excel, SQL, Python, and emerging technologies like ChatGPT. Their statistics calculators provide interactive tools that help students understand the mechanics behind statistical calculations, bridging the gap between theoretical knowledge and practical application.
Udemy's marketplace model supports this diversity by hosting over 100,000 courses, including many free options that allow instructors to share expertise with global audiences. The platform's filtering capabilities enable learners to identify resources that match their specific needs and learning preferences.
Industry Integration and Career Development
Major technology companies have recognised the value of contributing to global education initiatives, with Google, Microsoft and Amazon offering professional-grade courses at no cost. Google's Data Analytics Professional Certificate exemplifies this trend, providing industry-recognised credentials that directly align with employment requirements at leading technology firms.
These industry partnerships ensure that course content remains current with rapidly evolving technological landscapes, whilst providing students with credentials that carry weight in hiring decisions. The integration of real-world projects and case studies helps bridge the gap between academic learning and professional application.
The comprehensive nature of these educational opportunities reflects the complex requirements of modern data and technology roles. Successful professionals must combine technical proficiency with communication skills, statistical understanding with programming capability, and theoretical knowledge with practical application. The diversity of available courses enables learners to develop these multifaceted skill sets according to their career goals and learning preferences.
As technology continues to reshape industries and create new professional opportunities, access to high-quality education becomes increasingly critical. These courses represent more than mere skill development; they provide pathways for career transformation and professional advancement that transcend traditional educational barriers. Whether pursuing data analysis, software development, or artificial intelligence applications, learners can now access world-class education that was previously available only through expensive university programmes or exclusive corporate training initiatives.
The future of professional development lies in this combination of accessibility, quality, and relevance that characterises the modern online education landscape. These resources enable individuals to build expertise that matches industry demands, also maintaining the flexibility to learn at their own pace and according to their specific circumstances and goals.
Moves to Hugo
30th November 2022
What amazes me is how things can become more complicated over time. As long as you knew HTML, CSS and JavaScript, building a website was not as onerous as long as web browsers played ball with it. Since then, things have got easier to use but more complex at the same time. One example is WordPress: in the early days, themes were much simpler than they are now. The web also has got more insecure over time, and that adds to complexity as well. It sometimes feels as if there is a choice to make between ease of use and simplicity.
It is against that background that I reassessed the technology that I was using on my public transport and Irish history websites. The former used WordPress, while the latter used Drupal. The irony was that the simpler website was using the more complex platform, so the act of going simpler probably was not before time. Alternatives to WordPress were being surveyed for the first of the pair, but none had quite the flexibility, pervasiveness and ease of use that WordPress offers.
There is another approach that has been gaining notice recently. One part of this is the use of Markdown for web publishing. This is a simple and distraction-free plain text format that can be transformed into something more readable. It sees usage in blogs hosted on GitHub, but also facilitates the generation of static websites. The clutter is absent for those who have no need of the Gutenberg Editor on WordPress.
With the content written in Markdown, it can be fed to a static website generator like Hugo. Using defined templates and fixed assets like CSS together with images and other static files, it can slot the content into HTML files very speedily since it is written in the Go programming language. Once you get acclimatised, there are no folder structures that cannot be used, so you get full flexibility in how you build out your website. Sitemaps and RSS feeds can be built at the same time, both using the same input as the HTML files.
In a nutshell, it automates what once needed manual effort used a code editor or a visual web page editor. The use of HTML snippets and layouts means that there is no necessity for hand-coding content, like there was at the start of the web. It also helps that Bootstrap can be built in using Node, so that gives a basis for any styling. Then, SCSS can take care of things, giving even more automation.
Given that there is no database involved in any of this, the required information has to be stored somewhere, and neither the Markdown content nor the layout files contain all that is needed. The main site configuration is defined in a single TOML file, and you can have a single one of these for every publishing destination; I have development and production servers, which makes this a very handy feature. Otherwise, every Markdown file needs a YAML header where titles, template references, publishing status and other similar information gets defined. The layouts then are linked to their components, and control logic and other advanced functionality can be added too.
Because static files are being created, it does mean that site searching and commenting, or contact pages cannot work like they would on a dynamic web platform. Often, external services are plugged in using JavaScript. One that I use for contact forms is Forminit. Then, Zapier has had its uses in using the RSS feed to tweet site updates on Twitter when new content gets added. Though I made different choices, Disqus can be used for comments and Algolia for site searching. Generally, though, you can find yourself needing to pay, particularly if you need to remove advertising or gain advanced features.
Some commenting service providers offer open source self-hosted options, but I found these difficult to set up and ended up not offering commenting at all. That was after I tried out Cactus Comments only to find that it was not discriminating between pages, so it showed the same comments everywhere. There are numerous alternatives like Remark42, Hyvor Talk, Commento, FastComments, Utterances, Isso, Mouthful, Muut and HyperComments but trying them all out was too time-consuming for what commenting was worth to me. It also explains why some static websites even send readers to Twitter if they have something to say, though I have not followed this way of working.
For searching, I added a JavaScript/JSON self-hosted component to the transport website, and it works well. However, it adds to the size of what a browser needs to download. That is not a major issue for desktop browsers, but the situation with mobile browsers is such that it has a sizeable effect. Testing with PageSpeed and Lighthouse highlighted this, even if I left things as they are. The solution works well in any case.
One thing that I have yet to work out is how to edit or add content while away from home. Editing files using an SSH connection is as much a possibility as setting up a Hugo publishing setup on a laptop. After that, there is the question of using a tablet or phone, since content management systems make everything web based. These are points that I have yet to explore.
As is natural with a code-based solution, there is a learning curve with Hugo. Reading a book provided some orientation, and looking on the web resolved many conundrums. There is good documentation on the project website, while forum discussions turn up on many a web search. Following any research, there was next to nothing that could not be done in some way.
Migration of content takes some forethought and took quite a bit of time, though there was an opportunity to carry some housekeeping as well. The history website was small, so copying and pasting sufficed. For the transport website, I used Python to convert what was on the database into Markdown files before refining the result. That provided some automation, but left a lot of work to be done afterwards.
The results were satisfactory, and I like the associated simplicity and efficiency. That Hugo works so fast means that it can handle large websites, so it is scalable. The new Markdown method for content production is not problematical so far apart from the need to make it more portable, and it helps that I found a setup that works for me. This also avoids any potential dealbreakers that continued development of publishing platforms like WordPress or Drupal could bring. For the former, I hope to remain with the Classic Editor indefinitely, but now have another option in case things go too far.
Controlling clearance of /tmp on Linux systems
19th June 2015
While some may view the behaviour in a less favourable, I always have liked the way that Linux can clear its /tmp directory every time the system is restarted. The setting for this is in /etc/default/rcS and the associated line looks something like:
TMPTIME=0
The value of 0 means that the directory is flushed completely every time the system is restarted, but there are other options. A setting of -1 makes the directory behave like any other one on the system, where any file deletions are manual affairs. Using other positive integer values like 7 will specify the number of days that a file can stay in /tmp before it is removed.
What brought me to this topic was the observation that my main Linux Mint system was accumulating files in /tmp and the cause was the commenting out of the TMPTIME=0 line in /etc/default/rcS. This is not the case on Ubuntu, and using that is how I got accustomed to automatic file removal from /tmp in the first place.
All of this discussion so far has pertained to PC's where systems are turned off or restarted regularly. Things are different for servers of course and I have seen tools like tmpreaper and tmpwatch being given a mention. As if to prove that there is more than one way to do anything on Linux, shell scripting and cron remain an ever present fallback.
Piggybacking an Android Wi-Fi device off your Windows PC's internet connection
16th March 2013
One of the disadvantages of my Google/Asus Nexus 7 is that it needs a Wi-Fi connection to use. Most of the time, this is not a problem since I also have a Huawei mobile Wi-Fi hub from T-Mobile and this seems to work just about anywhere in the U.K. Away from the U.K. though, it won't work because roaming is not switched on for it and that may be no bad thing with the fees that could introduce. While my HTC Desire S could deputise, I need to watch costs with that too.
There's also the factor of download caps, and those apply both to the Huawei and to the HTC. Recently, I added Anquet's Outdoor Map Navigator (OMN) to my Nexus 7 through the Google Play Store for a fee of £7 and that allows access to any walking maps that I have bought from Anquet. However, those are large downloads, so the caps start to come into play. Frugality would help, but I began to look at other possibilities that make use of a laptop's Wi-Fi functionality.
Looking on the web, I found two options for this that work on Windows 7 (8 should be OK too): Connectify Hotspot and Virtual Router Manager. The first of these is commercial software, yet there is a Lite edition for those wanting to try it out; that it is not a time limited demo is not something that I can confirm though that did not seem to be the case since it looked as if only features were missing from it that you'd get if you paid for the Pro variant. The second option is an open source one that is free of charge apart from an invitation to donate to the project.
Though online tutorials show the usage of either of these to be straightforward, my experiences were not all that positive at the outset. In fact, there was something that I needed to do, which is how this post has come to exist at all. That happened even after the restart that Connectify Hotspot needed as part of its installation; it runs as a system service, which is why the restart was needed. In fact, it was Virtual Router Manager that told me what the issue was, and it needed no reboot. Neither did it cause network disconnection of a laptop like the Connectify offering did on me and that was the cause of its ejection from that system; limitations in favour of its paid addition aside, it may have the snazzier interface, but I'll take effective simplicity any day.
Using Virtual Router Manager turns out to be simple enough. It needs a network name (also known as an SSID), a password to restrict who accesses the network and the internet connection to be shared. In my case, the was Local Area Connection on the dropdown list. With all the required information entered, I was ready to start the router using the Start Network Router button. The text on this button changes to Stop Network Router when the hub is operational, or at least it should have done for me on the first time that I ran it. What I got instead was the following message:
The group or resource is not in the correct state to perform the requested operation.
While the above may not say all that much, it becomes more than ample information if you enter it into the likes of Google. Behind the scenes, Virtual Router Manager uses native Windows functionality to create a Wi-Fi hub from a PC, and it appears to be the Microsoft Virtual Wi-Fi Miniport Adapter from what I have seen. When I tried setting up an ad hoc Wi-Fi network from a laptop to the Nexus 7 using Windows' own network set up capability via its Control Panel, it didn't do what I needed, so there might be something that third party software can do. So, the interesting thing about the solution to my Virtual Router Manager problem was that it needed me to delve into the innards of Windows a little.
Firstly, there's running Command Prompt (All Programs > Accessories) from the Start Menu with Administrator privileges. It helps here if the account with which you log into Windows is in the Administrator group, since all you have to do then is right-click on the Start Menu entry and choose Run as administrator entry in the pop-up context menu. With a command line window now open, you then need to issue the following command:
netsh wlan set hostednetwork mode=allow ssid=[network name] key=[password] keyUsage=persistent
When that had done its thing, Virtual Router Manager worked without a hitch though it did turn itself after a while and that may be no bad thing from the security standpoint. On the Android side, it was a matter of going in Settings > Wi-Fi and choosing the new network that was created on the laptop. This sort of thing may apply to other types of tablet (Dare I mention iPads?) so you could connect anything to the hub without needing to do any more on the Windows side.
For those wanting to know what's going on behind the scenes on Windows, there's a useful tutorial on Instructables that shows what third party software is saving you from having to do. Even if I never go down the more DIY route, I probably have saved myself having to buy a mobile Wi-Fi hub for any trips to Éire. For now, the Irish 3G dongle that I already have should be enough.