Open-source alternative
Software defined by the proprietary product it replaces: the open-source Notion, the open-source Salesforce, the open-source Photoshop. The dominant way this category is described, sold and discovered — and a frame that shapes what gets built, not only how it is marketed.
Why the frame wins
It answers the question people actually arrive with. Someone looking to leave a subscription knows the product they want to leave, and has no vocabulary for the software that might replace it. "Open-source alternative to X" turns an unfamiliar landscape into a lookup. openalternative exists because that lookup is valuable enough to fund with sponsorships: 500+ projects across 20+ categories, all filed against something commercial.
The frame also does real work for a project. It supplies a feature list to be measured against, an audience already sold on the use case, and a one-line description that needs no explanation.
What it distorts
It ranks by recognition. openalternative orders entries by GitHub stars and nothing else — no maintenance signal, no release recency, no security history. Stars accumulate at the moment of discovery, which favours software that is easy to describe in one line. ever-gauzy — nine business modules, 27,198 commits — sits below a list of links on that measure.
It has no slot for software with no commercial twin. proton-wine-andrevto replaces nothing; it makes Windows games run on Android phones. omniwm is not an alternative to a product, it is a window manager for a platform that ships no such thing. Neither can be filed in a directory organised by what a project is an alternative to, and both are load-bearing for the people who use them. The frame is a discovery mechanism that systematically underweights compatibility work, infrastructure, and anything whose value is that it exists rather than that it is cheaper.
It sets the roadmap by imitation. A project measured against a commercial product is under pressure to match that product's feature list, including the parts that exist for the vendor's business reasons rather than the user's.
The clearest case of that, added 2026-08-30. alexandrie's own repository description names three products before it says what the software is — "The open-source, offline-first Notion, Obsidian & Confluence alternative" — and the feature list can be read straight off them: nested documents, Kanban boards and a Ctrl+K palette from Notion; teams, workspaces, five permission levels and OIDC/SSO from Confluence; markdown-first authoring and offline use from Obsidian. Three targets, three inherited roadmaps, assembled into one product.
It also shows where the imitation stops. Obsidian's actual proposition is a folder of markdown files you can open with anything; Alexandrie stores documents in MySQL and offers a ZIP export instead. So the frame carried across the feature that is easy to name (markdown) and not the one that does the work (files on your disk). A project can be an alternative to a product's description without being an alternative to its architecture, and the description is what the frame transmits.
openpanel is the first instance to look at, because it is the only source that applies the frame to itself: "an open-source alternative to Mixpanel" is its own GitHub description, and its feature list — session replay, A/B testing, funnels — is Mixpanel's product surface. One case, and an ambiguous one, since those features are also what anyone shopping for analytics asks for. The hypothesis has evidence now and is still not a finding.
A directory built on a different axis sees a different set. awesome-terminals sorts 403 entries by where a thing runs and what it was written in. Neither axis asks what anything replaces, so neither of the two projects that fall through the gap above could fall through it there. It pays for that immediately: a category axis admits everything in the category — iTerm2 and a dead commercial emulator sit beside Alacritty — so that list is useful here without being evidence about open source at all. The choice of axis, not the choice of metric, is what decides what a directory can see.
The frame turned sideways (added 2026-09-05). locally-uncensored runs the comparison table against Open WebUI, LM Studio, Jan and SillyTavern — four FOSS applications, two of them AGPL like it is — and links eight "vs" posts on its own domain. There is no proprietary incumbent in the column at all. What that table sells is not freedom from a vendor but consolidation: one installer replacing four things the reader already could have had for nothing. It is the clearest evidence yet that the frame is a marketing axis rather than a claim about ownership, because here it is doing work with the ownership question removed.
The alternative framing this spoke prefers
foss-application organises by who is in control rather than by what is being replaced. That frame keeps compatibility layers, environment software and self-hosted services in the same category where they belong, and it does not quietly rank a piece of software by how famous the thing it imitates is.