field notes

Your bytes are your bytes: what leaving actually costs

Every storage product runs two meters: the monthly one on the pricing page, and the exit meter you only read the day you leave. The second one tells you who owns the library.

Leland Dutcher · JuiceMount's author published 7 min read

Competitor facts checked June 2026; verify before relying on them.

Two meters run on every storage product you will ever use. The first is the monthly one, printed on the pricing page in a large friendly font, and everyone reads it. The second is the exit meter: what it costs, in dollars and days, to take the whole library somewhere else. Nobody reads that one until the day they need it, which is the day they discover whether they own the library or rent access to it. This post reads the second meter in public, with sources, and is honest about where our own project's interest lies: we build JuiceMount, and the exit story is one of its arguments.

the thesis, drawn · two meters per product

meter 01 · monthly

reads $/mo, budgeted

printed on the pricing page in a large friendly font, read every month

meter 02 · exit

reads $ + days, once

unread until the day you leave; it tells you who owns the library

Two meters, one product. The monthly meter is familiar, budgeted, and read every month; the exit meter sits dark until the day you leave, and that day it is the only number that matters. No prices drawn on purpose: the rest of this post reads the second meter with sources.

Pulling 20 TB out, priced #

Take a 20 TB library, a normal size for a small post team, and price the one-time pull at published rates, checked June 2026 (1 TB = 1,000 GB here):

the exit bill · pulling 20 tb out

AWS S3100 GB free, then 9,900 GB × $0.09 + 10,000 GB × $0.085 ≈ $1,741
Backblaze B2 leaving is free $01
Cloudflare R2 leaving is free $02
Wasabi leaving is free $03

checked june 2026 · sources below

One-time pull at published rates, 1 TB = 1,000 GB. 1 egress free up to 3× stored volume per month, so a 20 TB library has 60 TB of monthly headroom; $0.01/GB beyond that. 2 zero egress at any volume, by policy. 3 free under their reasonable-use policy, and a one-time migration qualifies; deleting young data can trigger their 90-day minimum-storage charge, a separate line worth knowing about.

Same bytes, same wire, a $1,741 spread, decided entirely by which logo is on the bucket. The same four rows run live in the priced exit on the compare page. And if your library lives in MinIO on your own NAS, this bill does not apply at all: copying out crosses your own LAN and no egress bill exists.

The vendors, named fairly #

The storage SaaS tools for editors sit a layer above those rates, and they deserve precise language rather than scary language, all checked June 2026. Their full published prices sit side by side in the compare page's lane tables, and they are weighed as working tools in the LucidLink alternatives rundown; here the only question is the exit.

iconik is the explicit one: egress is billed through its prepaid credits, per its own cost documentation. Getting media out of iconik-managed storage costs real money, quantified only through sales.

LucidLink documents no separate egress or exit fee; their own writing says storage and streaming are included in one monthly price with no separate egress line items. The precise phrase is none documented, not free: there is also no documented bulk-export tool, so leaving is a re-download project you run yourself through the client, at your line's speed, while the subscription stays alive long enough to finish it.

Shade likewise documents no egress, export, or exit fee anywhere public; their billing docs speak only of seats and storage. Same phrase, same reason: none documented, and the same re-download project on the way out.

Suite on the managed plan documents no egress fee either. On their bring-your-own-storage plan the answer is structural: the bucket is yours at a cloud provider, Suite's own terms say the provider bills you separately, so the exit costs whatever your provider's egress rate says. BYO on R2 exits at $0; BYO on S3 exits at the bill above.

True even at $0: the wire #

Suppose the meter reads zero. The bytes still have to move, and 20 TB at a sustained 1 Gbit/s is about 2 days of wire time, whoever you are leaving. That number is plain arithmetic, but it changes what the exit meter means: the question is not only what leaving costs, it is whether a monthly bill keeps running while the wire does its two days of work, and whether anything in the contract is in a hurry on your behalf. An exit you can run calmly, off your own hardware, on your own schedule, is a different project from one running against a renewal date. Wire physics has its own post: why your NAS feels slow over a VPN counts what latency, rather than bandwidth, does to a working day.

the wire, drawn · 20 tb at 1 gbit/s

≈ 2 days of sustained transfer, whoever you are leaving

20 tb at a sustained 1 gbit/s · plain arithmetic

The strip is two days wide and the fill stops just short of the second tick. It fills at the same speed whoever you are leaving; the meter question is whether a bill or a deadline runs while it does.

Policy versus architecture #

Everything above is policy: published prices and terms, true the day we checked them and changeable by their owners any day after. This is not a hypothetical. Shade's published Growth price in June 2026 is $35 monthly, $29.75 annual; TechCrunch's coverage of their funding round, published 2026-04-22, described a $20 per seat per month plan with the same 500 GB of active storage. We cannot prove a before-and-after from the outside, so here is the supportable weaker claim: within one quarter, the price on the public record moved, and it did not move down. LucidLink's published $27 is itself labeled a promotion off a $32 list price, checked June 2026: the price you sign at and the price on the list are already two different numbers. None of this is scandal. It is what published prices do, and a policy that says leaving is free today is the same kind of object.

Architecture is a different kind of object. An on-disk format that is open and documented cannot be repriced, because there is no one to send the new bill. The promise is not in a terms document; it is a property of the bytes.

Ownership at the architecture level #

This is the level JuiceMount aims at, and the claim is checkable rather than promised. File data sits in a bucket you control, on your NAS or at a provider you chose, in JuiceFS's open, documented chunk format. The volume is a standard JuiceFS volume, formatted by the stock tooling, and the README states the consequence plainly: the stock juicefs client can mount it with no JuiceMount involved. If this project disappeared tomorrow, your volume would not notice.

the open format · mountable without us

The exit that cannot be repriced, drawn. Per the project README, the volume is "a standard JuiceFS volume", formatted with stock tooling, "so the stock juicefs client can mount it with no JuiceMount involved": cross this project out and the volume still opens.

Leaving is correspondingly boring, which is the point. Copy files off the mounted volume, or juicefs sync the tree out: the same migration-grade tooling that moves a library around is the exit tool, and the Manager web UI that migrated your existing library in at the start was the rehearsal. There is no exporter to request, no credits to buy, and the bucket never stops being yours during any of it.

The honest counterweight #

Here is what the exit meter does not measure, and where the SaaS bill buys real things. A managed service brings managed durability, a support queue, and someone whose pager goes off at 2 AM who is not you. Self-hosting brings the opposite deal, and our README says it without flinching: a failed disk on your NAS is your failed disk, the 3-2-1 backup discipline is yours, and that is the deal that makes it free. If nobody on the team wants that deal, the monthly meter is not waste; it is the correct price of not carrying the pager. The argument of this post is narrower: whatever you pick, pick it knowing what both meters read.

The test to run before the first upload #

So here is the test, applicable to any vendor including us. Before the first byte goes in, ask what leaving costs in dollars and in days. If the answer is a published egress rate, you can price it; the calculator's exit line does that arithmetic at your own library size. If the answer is an open on-disk format on your own hardware, you can verify it. If the answer is a credit system, a quote, or a shrug, you have learned what you needed to know, cheaply, and earlier than most teams learn it.

next step

The compare page prices the exit interactively at the same published rates, and the calculator carries an exit line for your own library size next to the year-one math.

The exit, priced, on the compare page The calculator's exit line

Know a vendor's exit terms we should cite? Discuss on GitHub.

Where these numbers come from