User guide
Install
Download the latest release from GitHub Releases.
- Windows 10/11 (x86_64): run
routemap-<version>-windows-x86_64-setup.exe. Choose Install for me only (no administrator rights, installed in your user folder) or Install for all users (always Program Files). Route Map is then in the Start menu and in Settings > Apps; tick Add routemap-cli.exe to PATH to use the command line in any terminal.routemap.exenever opens a console;routemap-cli.exebeside it is the command line. Until the build is code-signed, SmartScreen may say "Windows protected your PC": choose More info, then Run anyway. Without installing: extractroutemap-<version>-windows-x86_64.zipand runroutemap.exein theRoute Mapfolder. - macOS 12+: open the
.dmgfor your Mac (macos-arm64for Apple silicon,macos-x86_64for Intel) and drag Route Map to Applications. The app is not notarised, so the first start is blocked: right-click it, choose Open, then Open again (on macOS 15: System Settings > Privacy & Security > Open Anyway). - Linux x86_64:
sudo apt install ./routemap-<version>-linux-x86_64.deb(Debian, Ubuntu and relatives) orsudo dnf install ./routemap-<version>-linux-x86_64.rpm(Fedora, RHEL;zypper installon openSUSE). Route Map is then in the applications menu androutemapon the PATH. Without installing:chmod +xthe.AppImageand run it, or unpack the.tar.gzand run./routemap. Where the distribution does not allow unprivileged ICMP, traces usetraceroute, which then has to be installed.
Upgrade by installing the new version over the old one (installer, package or DMG); settings, history and databases stay. Help > Check for Updates says whether a newer version is out and offers the file for your system. Download and check fetches it from this project's GitHub releases into a folder only you can read and checks three things before anything runs: that SHA256SUMS.ed25519 is a signature by the update key built into this version over the release tag and SHA256SUMS, that SHA256SUMS lists the file once, and that the file's SHA-256 matches. Then it runs the installer (Windows), opens the disk image (macOS), opens the .deb or .rpm in your software installer, or replaces a running AppImage in place and restarts. A failed check deletes the download and says which check failed. These downloads go to GitHub directly, so they do not appear in the project site's download statistics. Versions up to 0.2.0-beta.5 open the download in the browser instead. Uninstall: Settings > Apps on Windows, the Trash on macOS, sudo apt remove routemap or sudo dnf remove routemap on Linux. Settings and history are kept in %APPDATA%\routemap, ~/Library/Application Support/routemap or ~/.config/routemap until you delete that folder.
Check a download (optional): import the release key (RELEASE-KEY.asc) with gpg --import RELEASE-KEY.asc, run gpg --verify SHA256SUMS.asc SHA256SUMS, then sha256sum -c SHA256SUMS --ignore-missing (Linux), shasum -a 256 -c SHA256SUMS --ignore-missing (macOS) or Get-FileHash <file> (PowerShell). The release key is D57C 7E26 C19F 9436 E2D6 6F37 4080 97D1 91DD F981.
First trace
On the first start a short tour points at the origin, the target field, the map, the hop table and the menus; Skip tour or Esc ends it, the arrow keys move through it, and Help > Show the Tour brings it back. Settings is Route Map > Settings on macOS and Edit > Settings on Windows and Linux.
- Type a hostname or IP address in the box at the top and press Trace.
- Hops appear on the map and in the table as the trace runs. The tool's own output is under Tool output below the table (View > Tool Output).
- When it finishes, the map fits the route. Zoom with the wheel or pinch, drag to pan, Fit to see the route again, World for the whole map.
Your starting point (the origin) comes from your public IP address until you set one in Settings > Origin: a city, coordinates, or a point picked on the map. A set origin is remembered, and while one is set the public-IP lookup never happens.
Reading the result
Every hop row says where its location came from, in the Source column, and the marker has the same colour:
| Source | Meaning |
|---|---|
hoiho | The router's hostname matched one of CAIDA Hoiho's rules for that operator, and the rule names the city. |
site-code | The hostname contains a carrier's own site code (for example ffm for Frankfurt on Arelion's network), from the bundled table. |
ip-db (DB-IP) | Neither hostname source knew it, so the offline DB-IP Lite City database on this machine placed it. Treat it as a hint: backbone addresses often geolocate to where the prefix was registered. |
ip-db (RIPEstat, online) | The same, answered by RIPEstat online, because DB-IP Lite City is not installed or did not know the address. |
ip-db, country only | The database knew only the country. The marker is hollow, sits in the country, and the map does not zoom out for it. |
anycast, served from near X | The destination only. It answered faster than the place the database registers it at allows (bbc.com's Fastly address, registered in Montreal, answers from Manila in 5 ms), so it is served from a node near you: placed at the largest city reachable in the RTT it added past the last placed hop before it, and the summary says "likely anycast". An inference, not a measured place. |
inferred from neighbours | No source placed it (no hostname rule; the database's place is ruled out by the RTT), but the hops either side are placed in one city, so it is shown there with a hollow marker. An inference from the path, not an answer about the router: Arelion's link to Plus.line (62.115.153.153, plusline-ic-...twelve99-cust.net) between two Frankfurt hops. |
local | Your own network or your ISP's carrier NAT. Never looked up anywhere. |
unresolved | Not placed: it did not answer, or no source placed it within the physics bound. The Unplaced panel says why. |
Every placement, whatever its source, is checked against the round-trip time. Light in fibre covers about 100 km per millisecond of round trip, so a hop answering in 5 ms cannot be more than about 500 km away (plus a small allowance for the origin being approximate). A placement farther than that is rejected and the row's Notes say so. A database placement between two hops in the same area is also rejected when the RTT did not rise enough to pay for the detour (a Frankfurt hop, then "Chicago", then Frankfurt again 46 ms later).
Hops in the same city in a row collapse into one marker (for example 14-15). Click a marker to select its rows; click a row, or use the arrow keys, to highlight its marker. Esc clears. Hover a marker to see its row's fields.
The Notes column also flags what the trace itself shows: a hop that answers only some probes because the router rate-limits ICMP (not real loss further on), a destination that does not answer ICMP, and RTTs that jump in a way that suggests the reply took a different path back.
Loss at a hop counts as ICMP rate limiting when any later hop that answered shows less of it: the packets were getting through, and the router only declined to answer some probes. Its Loss cell is greyed. The route summary, the PDF and the JSON export ("loss") report only the loss measured at the destination; when the destination never answers, its loss is reported as unknown, never as 0%.
The map
Lines follow the great circle between hops, the way the packets actually fly, and are coloured by how much round-trip time each step added: grey under 15 ms, warming to the hot colour at 60 ms (both set in Settings > Map). A dashed line crosses hops that did not answer or a country-only placement.
Settings > Map > Colours has a Colour-blind safe option: rust to navy on the light map and yellow to vermilion on the dark one, chosen with a colour-vision simulation so that the three steps stay apart for protan, deutan and tritan colour vision, each at least 3:1 against the map. Hot steps are also drawn dash-dot, so colour is never the only cue. The exports use the same colours.
Traces probe the same way on every platform: ICMP echo from the app's own prober, 30 hops, three probes per hop, one second per reply, with every router that answers a hop listed (a hop answered by two routers shows both). Settings > Trace offers the system traceroute for UDP probes, or TCP probes with administrator rights.
View > Theme (or Settings > Map) sets System, Light or Dark for the window, the map and the exports.
Flat and Globe switch projection (View > Globe, Ctrl+G); the globe is centred on the route and turns when you drag it.
Moving around the map:
| Flat map | Globe | |
|---|---|---|
| Mouse wheel | zoom around the pointer | zoom |
| Trackpad pinch (or Ctrl + scroll) | zoom around the fingers | zoom |
| Trackpad two-finger scroll | pan | turn |
| Drag | pan | turn |
| Double-click | zoom in one step | zoom in one step |
+ and - (after clicking the map) | zoom | zoom |
| Arrow keys | previous or next hop | previous or next hop |
Home and End | first hop, destination | first hop, destination |
Enter, Esc | select the hop's rows, clear | select the hop's rows, clear |
Shift + arrow keys | pan | turn |
0, or the Fit button | frame the route | centre on the route |
Tab moves between the target field, the map, the summary and the hop table; the arrow keys move between hops in the map and in the table, and each follows the other. Every control has a name for screen readers, and a hop is read out as its number, place, network, RTT, loss and notes.
Zoom stops at the whole world and at city-street level, markers and labels keep their size, and selecting a row centres its marker without changing the zoom. Zoomed in on the flat map, the coastline gets finer and city labels appear. Replay (Ctrl+R) draws the route again hop by hop.
The route summary
Above the hop table:
- AS path: the networks the trace crossed, from the bundled DB-IP Lite ASN database, with an RPKI badge per network (do the routes it announces have a valid ROA?).
- Countries transited, in order. Countries you list in Settings > Map > Sensitive countries are flagged.
- Destination: a "likely anycast" note when the destination answers too fast for where it is registered, or belongs to a large content network.
- Typical latency (RIPE Atlas): what RIPE's anchors measure between the two countries today, against what you measured.
- BGP view (RIPE RIS): how many of RIPE's route-collector peers see the destination prefix, and whether their AS paths match the one your packets took.
- BGP updates, last 48 h: routing activity for the destination prefix; a burst at trace time is called out.
- RTT per hop and the physics floor: measured RTT against the lowest RTT each placement allows.
The table gains ASN and RPKI columns. Select one hop for its details: address, how it was placed, AS, routed prefix, RIR, RPKI, abuse contact (with Copy) and an AS overview. Open in FalconEye (also on the right-click menu) opens IP Reputation in FalconEye with the address on the clipboard.
Everything in the summary that comes from RIPE needs Online lookups on; a source that does not answer shows "unavailable" and nothing else changes.
Compare two runs
Trace > Trace Again and Compare (Ctrl+Shift+R) runs the same target again and shows the earlier run faint and dashed under the new one, with changed hops ringed, the table rows tinted, and a summary such as "Marseille and Paris are gone (hops 8 to 9 then)". Compare with an Export compares the route on screen with a JSON export, and Compare with an Earlier Atlas Measurement with your own last RIPE Atlas traceroute to the target (no credits spent). Hops are matched by place, not by hop number, and a run that simply stopped answering is not reported as lost hops. The PDF includes the comparison.
Paths
Paths (next to Trace), Trace > Find Paths (Ctrl+Shift+T) or routemap TARGET --paths finds the paths a load balancer can send your packets along, each with its own loss and latency. Many networks spread traffic over several equal routes; an ordinary trace shows one of them, or a mix of them hop by hop.
How it works: Paris traceroute keeps every probe of one flow on one path (the fields routers hash stay the same while the probe's sequence number changes), and different flows can take different paths. At each hop Route Map adds flows until the multipath detection algorithm's stopping rule says, with 95% confidence, that no next hop is left unseen: 9 flows when a hop has one router, 17 when it has two, and so on. Each path found is then pinged ten times with one of its own flows, for its own latency and loss to the target.
What you see: the routes where the paths part ways drawn one colour per path, with the path's letter; a Paths column in the hop table and a sub-row for each router at a hop where the paths differ (3a, 3b); and the Paths list under the table with each path's share of the flows, the hop where it splits from the path listed before it (hover for every hop where it differs; the PDF and the JSON have them all), and its RTT to the target. When the target answered on some paths, a path whose ten attempts all went unanswered says 0 of 10 (the target is up; some servers answer fewer after the discovery's probes). When no path reached the target, each says not reached and names the last hop that answered. Choose a path in the list to show it alone on the map.
"At least": some load balancers do not spread every kind of probe, so a path can stay hidden; the count is a lower bound. A router that sends every packet another way (per-packet balancing) is named, and does not make paths of its own.
Limits, so it never floods a network: at most 20 probes a second (the default rate of CAIDA's scamper), at most 1,500 probes per discovery, the pings included (Settings > Trace can lower this), at most 64 flows, and it stops after 5 silent hops (scamper's and RIPE Atlas's default), then asks the target once more at full TTL on each path found, as RIPE Atlas does with its last probe. A path reached only by that last attempt is shown reached, with its silent stretch marked; "not reached" means the target was asked and answered on no path at all. The heading says why discovery ended ("stopped after 5 silent hops (3 to 7)"). A discovery takes one to three minutes: 1,500 probes at 20 a second is 75 seconds of sending, and each round of probes waits up to a second for the routers that do not answer. No administrator rights are needed anywhere.
Method. On macOS and Linux, Paths sends ICMP probes (ICMPv6 over IPv6). On Windows it sends TCP probes to port 443 instead, one connection attempt per probe, with the source port as the flow, and a router's answer is read from Windows itself (Windows 10 version 2004 or later). Each flow's source port is one Windows assigns, kept for the flow's later probes, so a port another program holds is never asked for; if another program takes a flow's port in between, the flow gets one new port and is probed again from hop 1. A probe this computer could not send is not counted as loss: the note under the Paths list says how many, and why (for example WSAEACCES 10013). Windows' own ICMP function picks each probe's identifier and sequence number, so it cannot keep a flow on one path; TCP can. A probe that reaches the target is a normal connection, closed at once. Load balancers spread TCP and ICMP by different fields, so the paths found can differ between the two; the method is shown in the Paths heading, the CLI output, the PDF, the PNG title and the JSON (route.paths.method: icmp-paris or tcp-paris, with port). Over TCP the last column of the Paths list is Target answered: of ten connection attempts per path, how many the target answered. They are spaced so no port is reused at once (a server can ignore a quick repeat from the same port), and a server can still ignore some, so a shortfall there is not necessarily loss on the path. The ten attempts are at least a second apart; some servers still answer fewer of them after the discovery's own probes, and the heading says so.
When TCP goes quiet (Windows). For each hop where no TCP probe got an answer, Paths sends one ICMP echo with the same TTL. If routers answer it, the heading says "Routers on this network do not answer TCP probes beyond hop N; Paths cannot map past it here; Trace (ICMP) can". If the target itself answers it, the heading says the target answers ICMP there but its TCP service only at full TTL, and that a firewall or load balancer in front of the target is likely.
The address. Paths, like Trace, resolves the target once and probes that address; the heading shows it. When the Trace before it resolved the same name to another address (a CDN can answer with a different one each time), both are shown.
DNS timing
Every trace times the lookup of its target: one lookup for IPv4 (A) and one for IPv6 (AAAA) through your system resolver, and the same through each resolver you add in Settings > Probes (none by default; IP addresses only). The summary under the map says which address was traced and how long the system resolver took ("bbc.com resolved to 151.101.0.81 in 18 ms"); the DNS section above the hop table lists every lookup with its time, its answer, whether the resolver said it is authoritative (the AA bit) and the answer's TTL. Neither your system resolver nor a recursive resolver says whether it answered from its cache, so that column reads "not stated". A computer without IPv6 gets no IPv6 address from the system resolver even when the name has one; the row says so and does not claim the record is missing. Each lookup gives up after 2 seconds, the whole step after 5.
The traced address is the one the trace was sent to: Route Map resolves the name once and gives the address to traceroute, tracert or mtr, so they cannot resolve it to another address of their own.
Plain DNS is not encrypted: the network between you and a resolver can see the name. Only your system resolver and the resolvers you list are asked; with Online lookups off (or --offline), only the system resolver, whose lookup the trace needs. Settings > Probes > Time the DNS lookup once per Live cycle adds one system-resolver lookup per cycle to continuous mode (off by default).
Port tests
Ports (next to Paths) or routemap TARGET --ports tries one TCP connection to each port in Settings > Probes (443, 80, 22, 25, 53 and 3389 by default) on the traced address and reports each as open (with the time to connect), closed (the target answered with a reset) or filtered (nothing answered within the wait, or an ICMP unreachable). For 443 and the ports listed under "TLS on", it also times a TLS handshake; certificates are not checked, so that time says nothing about the certificate. At most 32 ports, 4 at a time, single ports only: it is a reachability check for one target, not a port scanner, and ranges are refused. On Windows an unanswered port is given up after about a second (Windows sends one connection attempt and waits for its first retransmission timer); the result says so. Turn on Test these ports with every trace to add the test to each trace (off by default). The PNG never shows port results.
Next to a Paths run, the port test tells "not reached" apart: if port 443 answers, the target is up and the routers in between are what is silent.
Remote agents
A remote agent is a server you can already reach with ssh from this computer. Route Map runs one trace there and shows it with the server as its origin, so you can see the route from a data centre or another country.
On the server, install the engine for the user you log in as: pip install routemap-engine (Python 3.11 or later), then check it with routemap-engine agent 1 hello. A Windows server needs 0.8.1 or later for Paths, which fixed probes whose port another program held (pip install -U routemap-engine). In Settings > Agents, add the server by a name and its SSH destination (a Host from your ~/.ssh/config, or user@host) and choose where it is on the map: from the agent's public IP (a RIPEstat lookup made on the server, so RIPEstat sees the server's address; Route Map never shows or keeps it) or a city (no lookup). Test connection makes one connection and reports the engine version and what the server can probe.
Then choose the agent in the From: list next to Trace and click Trace, Paths or Ports. Route Map runs your own ssh once, with -T, BatchMode=yes and a 15 second connect timeout: an unknown host key or a key that needs a passphrase not loaded in ssh-agent fails at once instead of asking. It never retries and never connects in the background. Placing the hops and every lookup happen on your computer, with your Sources settings.
The first hops of a trace from a server are its provider's routers. They are hidden by default: listed, with no address, name or place, as are the server's own name and its gateway in the tool output. Show first hops in the agent's settings shows them. Live and reverse traces run from this computer only. Paths on a Linux server needs unprivileged ICMP sockets there (net.ipv4.ping_group_range); Route Map does not change a server's settings, and says so when Paths cannot run.
Continuous mode
Watch (next to Trace), Trace > Watch Continuously (Ctrl+Shift+W) or routemap --watch TARGET probes one target again and again, mtr style: every hop once per cycle, with the built-in ICMP prober and no administrator rights. The hop table switches to the mtr columns: Loss, Sent, Last, Avg, Best, Worst and StDev (the standard deviation of the answered RTTs, the jitter). The map is drawn from the first full cycle; later cycles only add hops that were not there before, so nothing is looked up again every second.
Under the table, the ping plot shows the selected hop (solid line) and the destination (dashed) over the last 5 minutes; the keys 1 to 4 choose 5 minutes, 15 minutes, an hour or the whole session. Lost probes are marks on the baseline, and a path change is a vertical rule with its cycle. The plot follows the row or marker you select and the RTT palette you chose in Settings.
Path changes are listed with their cycle and time: a hop answering from a different router, a hop starting or stopping to answer, the destination moving to another hop count. A hop's routers are compared over the last 10 cycles with the 10 before, so a load balancer alternating between the same routers is never reported as a change.
Loss follows the same rule as a single trace: loss at a hop that later hops do not share is ICMP rate limiting, greyed with its note; only the loss at the destination is real. If the computer sleeps or the app is suspended, the missed time is a gap in the plot and the statistics, never loss, and the session notes it. A change of the system clock (a time sync, or setting the clock by hand) is not a gap: the cycle is kept, the plot runs on without a jump, and the note says so, for example "Clock changed at 14:02:07 (+5 s); this cycle was kept." Route Map tells sleep from a clock change with a clock that keeps counting while the computer sleeps. Linux, macOS and Windows all have one; if it cannot be read, the note says "Clock check off" with the reason, and a clock change may then show as a gap.
Pause (P, when the cursor is not in a text field), Stop (Ctrl+.) and Reset counters (a button and a Trace menu item, with no shortcut) control a running session. A stopped session stays on screen: export it as PDF (the table, the changes, the gaps, the clock changes and one plot per hop for the whole session) or JSON (format version 5, with the statistics, the plot data, the changes, the gaps, clock_steps and clock_fallback), compare it with another run as with any trace, or reopen it from History or File > Open Trace.
Limits, in Settings > Live, so it never floods a network. They follow mtr:
| Default | Allowed | Where it comes from | |
|---|---|---|---|
| Interval | 1 s | 1 to 60 s | mtr's default, and the shortest it allows without root |
| Probes per second | at most 30 | fixed | what mtr sends by default on a 30-hop path |
| Wait for a reply | 1 s | fixed | as a single trace and mtr |
| Silent hops probed | 5 past the last answer | fixed | mtr's --max-unknown default in its manual |
| Session length | 1 hour | 5 minutes to 8 hours | a limit of Route Map's own |
Stopped sessions are kept in History, 50 MB of them at most together; past that the oldest sessions are dropped first. Continuous mode runs over IPv4 or IPv6, as Settings > Trace > IP version chooses, and never uses RIPE Atlas.
Paste a trace
File > Paste Trace (Ctrl+Shift+V) takes the output of traceroute, tracert or mtr --report run anywhere: another machine, a looking glass, a colleague's email. The dialog detects the format as you paste and asks where the trace was run from, because the origin anchors every physics check. File > Open Trace does the same for a saved file or a Route Map JSON export.
Exports
File > Export (Ctrl+E):
- PNG: the map with the legend, 1600 by 900, light or dark (Dark map).
- PDF: the map, the route summary (AS path, countries, RIPE data), the hop table with ASN and RPKI, any comparison, the unplaced hops and, if Include the raw trace text is on, the tool's output.
- JSON: everything, including the trace text, to open again later or to process elsewhere.
Exports are written only to the file you choose. The command line exports the same three formats; see the command-line reference.
Settings
- Origin: from your public IP address (one lookup, only while nothing else is set), a city, coordinates, or a point picked on the map.
- Trace: the tool (detected at startup:
tracerton Windows,tracerouteon macOS and Linux, ormtrif installed), its flags, and Give up after, the time limit for one trace. - Probes: the resolvers DNS timing also asks (IP addresses, none by default), Time the DNS lookup once per Live cycle, and the port test's Ports, TLS on and Wait per port, with Test these ports with every trace (off by default). See DNS timing and Port tests above.
- Agents: the servers you can trace from, with Add, Edit, Remove and Test connection. See Remote agents above.
- Sources: Online lookups switches every online source at once; off, nothing new leaves the machine. Under it, each source on its own: CAIDA Hoiho, the bundled carrier site-code table (offline), the IP geolocation database (RIPEstat), and reverse DNS for hops without a name. Keep Hoiho and IP database answers for sets the local cache lifetime (30 days by default); Clear cache empties every cache. Offline data shows the DB-IP Lite months installed, with Update now (download this month's) and Import database file (for a machine with no internet).
- Map: projection, the two RTT step colour thresholds, sensitive countries, and the FalconEye address used by Open in FalconEye.
- Trace also has IP version: Automatic (the address your system prefers), IPv4 only or IPv6 only, for a host that has both; it applies to traces, Paths and continuous mode. And the Paths probe budget (1,500 at most).
- RIPE Atlas: the API key and the Trace menu entry, and Reverse traces with Withdraw consent; see below.
- Privacy: the trace history (the last 50 traces, on by default, clearable).
Offline data
The app ships with DB-IP Lite ASN (for the AS path). On first start it offers DB-IP Lite City, about 60 MB, so IP database placements happen on your machine instead of at RIPEstat: Download, Import File (copy dbip-city-lite-YYYY-MM.mmdb.gz from https://db-ip.com/db/download/ip-to-city-lite on another machine) or Not Now. Until it is installed, the Source column says ip-db (RIPEstat, online). Both files are updated monthly by DB-IP; use Settings > Sources > Update now or routemap data update. Licence: CC BY 4.0, IP Geolocation by DB-IP.
RIPE Atlas mode
With a RIPE Atlas account you can trace from an Atlas probe on your own network instead of from this computer, which helps when your machine cannot run a trace (a locked-down laptop) or when you want a second vantage point.
- Create a key at atlas.ripe.net/keys with the schedule a new measurement permission, and credits read if you want your balance shown before each trace.
- In Settings > RIPE Atlas, paste it and switch on the Trace menu entry.
- Choose Trace > Trace from a RIPE Atlas Probe. The first time, confirm that the measurement will be public. Every time, the dialog shows your balance, what this trace costs and what is left after it. A key RIPE does not accept stops the trace there, with a button to Settings; a key without credits read hides the balance but can still trace.
Route Map finds a connected probe on your network's AS (or, failing that, in your country), schedules one traceroute (60 credits: RIPE charges a one-off measurement twice what a periodic one costs), and draws the result from the probe's location. Atlas measurements are public: RIPE NCC publishes every measurement, including the target. Your origin coordinates are never sent; they only rank probes on your machine.
Reverse trace
The route from you to a target is often not the route back. Reverse trace (next to the summary once a trace is on screen, or Trace > Reverse Trace via RIPE Atlas) asks a RIPE Atlas probe near the target to trace back to you, and shows both directions side by side.
- The probe is a connected one in the target's network (its AS), or failing that in its country, nearest the target, and never one in your own network. If there is none, nothing is scheduled and no credits are spent.
- It publishes your public IP address. Atlas measurements are public, and a reverse trace's target is your public IP. Before the first one Route Map says so and asks you to agree; nothing runs without that. Every time, it shows the probe, the IP address that will be published (it changes when you change networks) and the cost, 60 credits. Settings > RIPE Atlas > Withdraw consent stops reverse traces at once until you agree again.
- It usually takes 1 to 3 minutes, and longer when routers on the way back do not answer: Route Map asks the probe to wait 2 seconds for each of 3 replies at a hop (RIPE's default is 4), so a silent hop costs 6 seconds. A router that answers slower than 2 seconds counts as silent.
- The result is drawn dashed in its own colour, with diamonds where the two directions split and rejoin, and listed under the table aligned by network, not by hop number, with the rows that differ tinted. Different routes each way are normal on the Internet. When several hops in a row stay silent, Atlas sends one last probe with TTL 255; its answer is shown as the next hop with the note "TTL 255 probe", as RIPE's own results page numbers it.
- It is kept in History and in the JSON export with the trace it reverses, and the PDF has a Reverse trace section. From the command line:
routemap TARGET --reverse --publish-my-ip(the second option is the agreement, given for that run only).
Using Route Map behind a VPN or SASE client
A VPN or SASE client changes what a trace can see. Route Map tells you when it can detect that, but what it measures is still the path your packets took.
Why a trace can mislead here
- The origin can be wrong. On Auto, your origin comes from your public IP address. Behind a tunnel, that is usually the provider's exit, not where you are. RTT checks are anchored on the origin, so correct hops can be rejected and wrong ones accepted. Set your real location in Settings > Origin.
- TCP can be intercepted. Some clients proxy TCP on port 443, so the "target" that answers is a local or nearby proxy. Paths on Windows probes TCP 443, so it can be affected. RTTs and hops from such a run mean little.
- Hops can be hidden. The path between this computer and the provider's entry point can collapse into one or two hops, or none, and tunnel overhead adds to every RTT.
- ICMP and TCP can take different routes. Depending on the client's policy, a Trace (ICMP) and Paths (TCP on Windows) can measure different routes in the same session.
What the notices mean
- "Traffic to this target goes through [adapter]." The route to this target uses a tunnel adapter. The name is the one your system reports. The map shows the path through it, not the direct internet path.
- The origin warning appears when the origin is on Auto and your system's default route, so your internet traffic, goes through a tunnel. A tunnel that carries only some targets does not change your public IP and gets no warning. Route Map does not change the origin for you.
- "TCP results are unreliable." A TCP connection completed or was refused at TTL 2 or less, where a real probe must die at a router. It does not count if the target is on your local subnet, or if an ICMP echo reaches it at the same TTL. The export names the condition that triggered it.
- "TCP and ICMP left by different first hops." Shown only when both protocols got an answer from hop 1 and the answers differ.
An adapter your system does not clearly describe is listed as "unclassified". It is named in the Interface section of the route summary and in exports, with no notice over the result.
The probe interface
Settings > Trace > Probe interface lists the adapters your system reports, with their kind, state and addresses. Automatic (system route) is the default. On the command line, use -i NAME and --list-interfaces.
It applies to Trace, Paths and Watch. DNS timing and port tests keep the system route, and an agent probes from its own network.
Limits:
- Some clients intercept traffic in their driver and ignore which adapter a program asks for. After a run on a chosen interface, hop 1 is compared with that adapter. If hop 1 belongs to another adapter, the result says "Probes were redirected by [adapter]; the selected interface was not used." If it cannot tell, the result says so. Neither case is presented as a direct path.
- A connected VPN can disable a physical adapter's default route. The run then stops with a plain "no route" message, and nothing is sent through any other adapter.
- Choosing a physical adapter while a tunnel is up may bypass your organisation's security path. Route Map tells you once, and it is your choice. It never selects an interface by itself.
Seeing the real internet path
A trace from inside the tunnel cannot show the path outside it. Two things can:
- A remote agent (Settings > Agents) runs the trace from a server outside the tunnel.
- A RIPE Atlas reverse trace measures from a probe near the target back toward you.
Privacy
The adapter list, addresses and gateways are read locally and never sent anywhere. Exports carry the adapter's kind and name. Addresses and gateways go into an export only when you tick the checkbox in the Export dialog.
Privacy
Nothing is sent anywhere until you run or open a trace, and then only what PRIVACY.md lists (also Help > Privacy). Each source can be switched off in Settings; with all of them off, placements come only from the bundled carrier site codes.
Reporting a problem
Please open an issue on GitHub. If you would rather not use GitHub, email [email protected]. What helps most: your OS and version, the Route Map version (Help > About), what you traced, what you expected, and a screenshot or a JSON export (File > Export). A JSON export contains the trace, so leave it out if the path is private.
Help > Create Bug Report makes a zip with what helps most and nothing private by default: about.txt (Route Map and engine version and commit, OS, Python, Qt, kind of build) and settings.json with every key and token replaced by (removed), and your set origin (city or coordinates) and a FalconEye address other than the public one removed too. Tick Include the last trace to add the trace on screen as last-trace.json (target, every hop and your origin; the origin then stays in settings.json as well). The dialog shows every file exactly as it will be saved. Route Map sends nothing: you attach the zip yourself.
See also: troubleshooting, known limitations, questions.