HERMES is a Canadian company established in 2018 to research and develop a worthy successor to the cult-classic Eudora 7 mail client developed by QUALCOMM. Notwithstanding the coronavirus pandemic, we persisted, eventually expanding our offerings to encompass a Gnutella and BitTorrent servent (combined server and client) in Java, a computer game in Elixir, and a text-editing toolkit in C.
Due to circumstances beyond our control, it has been impossible to purchase or download any of our products since approximately September 2025, for new customers only. That constellation of issues has since been resolved and orders are possible again. If you are a new customer, please direct your purchase here; our previous order page is no longer active. For a retrospective of the factors that led to our extended unavailability, as well as some insights on how this can be prevented in future, see this link.
NOTE: Eudoramail 8.0 is sold through a pay-what-you-want scheme; it can be yours for as little as a dollar (we suggest an inflation-adjusted price of $90). Purchase it here by sending “support” for any amount (though this is not a donation—it’s a contract of sale, subject to all applicable laws in the Dominion of Canada). Be aware that the software is sent out in monthly batches and we are not responsible if you make your purchase and have to wait.
Our signature product, Eudoramail, is a rebuild of Eudora 7 for Windows as written by QUALCOMM, with language and security modifications to make it friendly for the modern, internationalised office environment, one where security is a paramount consideration.
If you don’t know what Eudora is, due to having grown up on the Internet post-2006, Tedium’s Ernie Smith has a fascinating write-up headlined “A Eulogy for Eudora” which, although it predates the release of our software in 2022, furnishes a useful retrospective on why QUALCOMM’s iteration of the software was ultimately pulled from the market and particularly on what made it so attractive to its customers.
In short, in the words of Steve Wozniak, in an interview he did with Lifehacker: Eudora “has an incredible feature that every single mail client should have.” He continues:
Any feature in the menu list, any action there, can be added as a button. I changed it so I have a vertical menu bar, so I can have tons and tons of pre-made buttons saved right where I want them up top, and I learn where those place are. You can script actions to the buttons, too, so I can quickly copy messages to my assistants. There are scripts I wrote for joke lists so I can forward a message, remove the brackets and formatting, and make sure all the original attachments are included, to a pre-defined ‘joke’ group. Apples Mail app just isnt scriptable enough to really handle my mail buttons.
Some of the buttons will re-direct mail with quote marks, or not. I’ve got another script that will actually customize a mail forward, like my own version of mail merge. So even if something’s going out to 400 people, I can set it to single out certain people and take away all the forwarding markings, so it looks like I singled out someone to send them mail. Which is, I hope, a nice little moment for them.
While the topic of security is too complex to discuss on this page, in general, eMail servers will deny connexion to a non-conforming client, rather than continuing with an insecure mode of communication. This caused some problems in the early days, before Eudoramail 8 even saw the light of day—problems such that the product would not be salable except to power users who were prepared to make the sacrifice of simplicity and ease of use. Worse, since then and continuing to the present, it has become obvious that electronic protective measures are not a fix-and-forget deal; they demand maintenance and ongoing revision. In short: security delayed is security denied!
The upshot of this is that, ever since May 22, 2018, when Computer History Museum curator Leonard J. Shustek released the Eudora™ Email Client Source Code, we have been shoring up Eudora’s EPM’s, even predating the launch of our product. According to Andrew Orlowski at The Register: “Under this latest deal, Qualcomm is to donate all IP – copyright code, trademarks and domain names – over to the museum. The code is released under a BSD licence with an agreement to bundle the binary of RogueWave’s 20-year-old MFC library (for non-commercial use).”
It may be remarked that Eudoramail 8.0 had a protracted development cycle by most people’s standards: “clean compilation of the main Eudora/Hermes executable” (but without linking to libraries) was achieved on 19 July 2019, while 31st March 2022 saw “source code that compiles, links with the Stingray library, runs, and does not immediately crash”. The initial public prototype (or alpha) was distributed on 1 August 2022, with at least forty-two subsequent alphas distributed since then.
Although originally intended as an open-source project in the conventional sense (with everyone being able to participate in development), there are serious technical reasons that this is not the case, and potentially will not be in the medium-term; instead, it might best be described as fair-source (i.e. source-available with caveats). Mr Orlowski highlights one of these reasons: compilation relies on Stingray, which went through a whole chain of acquisitions—first Rogue Wave, then QuovadX, and finally Perforce—and the binary is not enough; source code access is required. Compiling Eudoramail also demands Wintertree Software’s Sentry Spelling Checker Engine; although this is available as a dynamic-link library, our sense of ethics as an organisation will not allow us to pursue any route other than full source-code access. Finally, Eudoramail also relies on HERMES Paige—but that doesn’t really count, since it’s our rich-text library, and we produce it as an Open Source project in its own right. In short, Perforce prices Stingray’s source code at $3,600 per person. Wintertree prices Sentry’s source code at $5,000 per person. Development demands periodic compilation, meaning that each developer would need these libraries. While Sentry is slated to be replaced with a free solution, it would be contrary to all common sense to pursue the Open Source workflow for Eudoramail at the present time.
NOTE: Eudoramail 8.0 is sold through a pay-what-you-want scheme; it can be yours for as little as a dollar (we suggest an inflation-adjusted price of £55 Br., $70 Am., or $100 Cdn.). Purchase it here by sending a “donation” for any amount (though this is not really a donation—it’s a contract of sale, subject to all applicable laws in the Dominion of Canada). Be aware that the software is sent out in monthly batches and we are not responsible if you make your purchase and have to wait.
As of 2 April 2026, we finally did it: a compilable, fully Open Source successor to Eudora 6.2 for Mac is now in the wild. For more than one reason, this is a big deal, subject to certain caveats (about which more later).
Firstly, archival: older Macintosh apps used a feature called a resource fork, which included, at a minimum, things like text strings and icons. When transmitting Classic Mac source code over the Internet or on flash drive, resource forks could get corrupted or destroyed, leading to an inability to compile the product at all—unless the code was bundled using an archiver that knew about resource forks, such as StuffIt or BinHex. Tarballs (as common on later, UNIX-based Macs) or WinZip would strip resource forks deterministically and by design. The Eudora for Mac source code came to us with all the resource forks accidentally dummied out and unrecoverable, without significant effort. That effort has now been made.
Secondly, holes: the source code provided to us extended only to QUALCOMM’s intellectual property, not the intellectual property of others that they used with permission. This issue structurally parallels our experience with Eudora 7.1.0.9 for Windows (if anything, it’s neater) so we shan’t go into our solution of it here, but it has been solved (not rhetorically—factually).
Thirdly, toolchain: the QUALCOMM team historically used CodeWarrior as their integrated development environment, as common practice prior to 2004, while Apple development since then has almost universally relied heavily on Xcode. The master files that defined how the project was to be built had to be migrated.
In short, the process of establishing a reasonably modern build path demanded a vast amount of time and effort, which had to be abstracted from other parts of the project. The result is Eos 6.3, a Macintosh-first e-mail client that compiles on Xcode 9.4.1 under macOS X High Sierra and runs on Intel Macs running macOS X Mojave.
That said, this project does introduce a number of caveats, all tied to a number (!) of significant transitions undertaken by Apple in the last 20 years. Let’s unpack them.
With all of this in mind, the decision was taken to provide the new, yet obsolete Eos 6.3 to our loyal clients, along with all necessary support until such time as a final decision is made, hand-in-glove with stakeholder interests, on a roadmap charting out the form of the future Eudoramail for Macintosh. We are not contemplating the prospect of sunsetting the product at this time. Rather, we understand that product planning will necessarily involve choosing trade-offs, and this must not be done where the sun does not shine.
Partly because of the obsolescence of this codebase, and our interest in software preservation in a usable form, we have a fully open Eos repository. We make no claim on this project as our intellectual property; it is available strictly pro bono publico, although we do provide support and aim to use it as a springboard for our Macintosh product.
You may download the Eos codebase here and, if you wish, contribute to it.
Apps that manipulate text in some form are everywhere; this does not mean only “text editors” in sensu stricto, but also things like word processors, Web editors, Markdown editors, desktop publishing apps, and even the message composer embedded in your eMail client (which, on a technical level, is nothing more than a Web editor of a highly specialised form).
If you are producing an app in that (very broad) category, you need a text engine. If your background is in development for Windows, you will most likely be familiar with CRichEdit from the Microsoft Foundation Classes; if, by any chance, you do not use CRichEdit in your project, you are most likely writing your own text engine from scratch or using a cross-platform toolkit (Qt or GTK).
There is a fourth option: a turnkey text engine—one that is oven-ready right out of the starting gate—can offer a developer experience very much like CRichEdit, but with many more features thanks to its specialised nature. Our text engine, called Paige, can cope with almost any task you throw at it, from Unicode to mail merge, with a comprehensive API specification over 500 pages in length, available in PDF and HTML format.
Moreover, Paige’s API is reasonably high-level, so you don’t need to grapple with pointless mathematical minutiæ and can readily and rapidly progress to the “meat and potatoes” of document editing, aided by the time-honoured Cartesian coördinate system.
With Paige, the sky is the limit. While we’d never condone using it to produce an HTML viewer (that’s a horror story all on its own), the trick could undeniably be turned, and of course Paige could certainly cope with browsing Geminispace (the difference lies less in the nature of the task than in the quality).
HERMES Paige is available now from our Git repository and is fully Open Source, with a manual available in HTML and PDF formats. Get it today!
Arguably, WireShare was the first app we ever developed; in fact, it predates the founding of Team HERMES by many years, having seen its first version in 2010.
WireShare is a relatively difficult program to explain; in technical terms, it is a combined Gnutella and BitTorrent servent. Gnutella and BitTorrent are both ways for computers to transfer high-supply, high-demand files to each other. (Gnutella is older, but has certain advantages, such as a search function.) Because neither Gnutella nor BitTorrent make any distinctions between server and client, and people can and do serve files from their homes or offices, we call the type of file transfer that they do peer-to-peer. The advantage of this is much higher resilience; it’s naturally harder to take down a file that’s served from two hundred locations at once.
The problem, such as it is, is that peer-to-peer file sharing is abusable, and poor stewardship exacerbates any abuse. Prior to 2010, WireShare was known as LimeWire. Its corporate stewards, Lime Group LLC, were notorious for explicitly condoning abuse of the Gnutella network for violations of intellectual property law; inevitably, the Recording Industry Association of America took notice and filed suit, with the action being cited as Arista Records LLC v. Lime Group LLC.
During negotiations, Lime Group introduced a backdoor into their product, and actioned it (disabling recent—but not older—versions of their software) subsequent to U.S. Federal Court Justice Kimba Wood’s order. Fortunately, Team HERMES is not a U.S. company; it is proudly headquartered in Canada, and in any case, the terms of the consent decree only apply to Lime Group LLC and the software known as LimeWire. We distribute WireShare. Whether it is, or is not, functionally and graphically identical to LimeWire is immaterial.
HERMES WireShare can be downloaded at our Git repository here; source code and Java binaries are currently provided, and this will expand to pre-bundled binaries for most commodity operating systems (macOS, Windows, Haiku, and various Linux distributions) in the fullness of time.
The French board game of Trictrac is second only to Chess in its beauty and complexity; yet its equipment is among the most generic imaginable (a tables board as is used to play Backgammon, fifteen men per side, three coins, two Cribbage pegs, and a miniature flag) and its rules are, in a very real way, strictly algorithmic and computational—the phrase “clear, neat, and precise” can not but describe it perfectly. Unlike in partnership card games like Bridge or Quadrille, there is no “human factor” in Trictrac; nothing is negotiated, inferred, or socially mediated, and nothing is hidden, either. Every question or problem is soluble by resorting to the rules, and only the rules. This renders the game totally and utterly reducible to computer code (ideally, as it turns out, to event-driven Elixir code; never try to do this in JavaScript).
It is rather a pity, then, that until now, only two computer implementations existed, and even these were not ideal (one extremely old, and one forms part of a European Research Council superproject with an immense scope stretching to essentially all structured games in existence). We decided to fill this niche and introduce HERMES Trictrac, an online game server that supports networked play of six tables games: Backgammon (from the United Kingdom), Trictrac à écrire, classique, and combiné, as well as Toc (from France), and Toccategli (from Austria). To this we have added a number of smaller, simpler variants, as well as a few experiments, in the hope that they may offer at least a few hours of enjoyment but without any guarantees.
Five of our headline six games form a category of their own in that, unlike almost all other tables games, they are not race games, where winning is binary and means getting the playing pieces ‘home’ after a fashion (off the board); instead, they are point-accumulation games with plural (differential) scoring, in which winning means getting a target score and scoring means arranging and maintaining the men in one of several named configurations.
In case this sounds like impenetrable jargon, think of Rugby: there are multiple ways to score, each of which adds a different number of points to the total. In the five games here described, there are at least eight different ways to score, with varying point totals. These are subject to varying “wrappers”. For example, Trictrac classique is a bit like Lawn Tennis: twelve points make up one or two sets depending on whether a doubling flag is set to on or off, and winning twelve sets means winning the match.
The other game is Backgammon (i.e. the binary race game par excellence, with roots in Persia by way of England). Here, the rules are the same as that of the broader Tables family: one has fifteen men, one moves them to a specific place, from there one moves them off the board, and the first to be completely off the board wins.
This is all a long way of saying that, though they share a game board and pieces, binary race games and point-accumulation games have essentially nothing in common (structurally, as implemented in computer code), and we wish to do justice to the latter family while also leaving something novel for fans of the former: thus, we have Plakoto and Pheuga (from the Ottoman Empire) as well as a rotation game that consists of Backgammon, Plakoto, and Pheuga in order, all of which are easy to learn and extremely satisfying for a race-game lover, as well as the five games we already mentioned, which will doubtless put a smile on the face of any Trictrac devotee.
For some simple examples of our nonstandard race games: Plakoto (also known by its French name, Jacquet) has restricted movement (one piece, the Courier, must be borne off first), the pieces move in parallel (Black moves “backwards” compared to Backgammon), and there is no hitting. Pheuga (also known by its Bulgarian name, Tapa, or by its Arabic name, Mahbousa) has pinning in place of hitting (sitting on a blot prevents it from moving), and the pieces move in parallel. Tavli is Backgammon, Plakoto, Pheuga on loop until someone gets five, seven, or nine points.
That said, the canonical games we support are, and will ever remain, the national Tables games of Great Britain and of France: Backgammon, Trictrac classique, à écrire, combiné, Toc, and Toccategli.
HERMES Trictrac can be downloaded at our Git repository here. This is an online service for those who understand how to deploy an Elixir stack; we recommend fly.io and Gigalixir as the deployment services of choice. We have not yet deployed the service ourselves, though we plan to do this in short order. The rules of Trictrac (the game, rather than our implementation) can be found here in multilingual format (Danish, German, Swedish, and French).
UltraTerminal is a flexible, efficient client application for connecting PC users on Microsoft Windows with remote services. Other than limited legacy or sector-dependent use cases (see below), these services fall into two general buckets: for communication with mainframes (such as those running the IBM i operating system, as well as its predecessors OS/400 and i5/OS), the IBM 3270 and IBM 5250 standards are used, which deliver a screen at a time in a fill-in-the-blank (block mode) format much like that of paper forms; and for communication with UNIX and Linux servers, as well as more broadly, use is made of the HP VT100 and HP VT220 de facto standards, which are character-addressable (more like a text-only computer program).
Unlike some other apps in this market segment, we offer full support for telephone (traditional modem), telnet, and ssh connexions, without undue fussing with external software; this is particularly useful where ssh3270 and ssh5250 are concerned, due to the type of services that are often run through such links, as well as the lack of reasonably-priced alternatives that integrate ssh tunnelling.
UltraTerminal also supports, or plans to support, these other types of connexions:
The source code for HERMES UltraTerminal can be downloaded at our Git repository here; please take note, however, that the public source tree may, at times, appreciably lag with respect to the Windows binaries. We distribute the official version of the Windows binaries to our Eudoramail clients. If you wish to purchase, please do so by contributing a “donation” for any amount (though this is not really a donation—it’s a contract of sale, subject to all applicable laws in the Dominion of Canada). We suggest an inflation-adjusted price of £55 Br., $70 Am., or $100 Cdn. Be aware that the software is sent out in monthly batches and we are not responsible if you make your purchase and have to wait.