A connected press is also a network risk
# A connected press is also a network risk The press was ready and the press would not start. Plates were on. Paper was in the feeder. The minder had a signed sheet from the proofing round and a job he had run before. What he didn't have, on Friday afternoon, was the workflow server
A connected press is also a network risk
The press was ready and the press would not start. Plates were on. Paper was in the feeder. The minder had a signed sheet from the proofing round and a job he had run before. What he didn't have, on Friday afternoon, was the workflow server in the office, which had stopped after a power cut in that corner of the building, or the line out to the vendor, which had gone with it. The console wanted a handshake before it would let the job open. The handshake had nowhere to go. So the newest, most helpful machine in the hall stood still, while the older press beside it, the one nobody had got round to connecting properly, carried on printing. The connection had been sold as safety and speed. On that Friday it was a lock, and the key was in a room with the lights out.
A connected press is a better press when the links are up and the people are clear about authority. It's also a new way to be stopped, and a new way for a setting to change without the crew understanding why. You don't need a villain in this story. You need a server, a vendor line, and a habit of letting them become required. The practical risks are ordinary: who is allowed to change a setting, what a dead link does to a Friday job, and whether the plant can still run when the outside world has nothing useful to say.
Three links, and none of them is the feeder
Think of three connections, because they fail differently and they deserve different rules. The press itself: the console, the sensors, the keys or the digital front end, the bits of software that sit in the machine and will keep working if you leave them alone. The workflow server: tickets, files, presets, the schedule, often a box in the office or a service you reach over the internal network. The vendor connection: remote support, a licence check, a queue that lives outside the building, an update someone else decides is due. Helpful things ride on all three. So do dependencies you didn't mean to create.
The press link is the one the crew can see. A preset arrives, a curve updates, a job appears in the queue. When this works, makeready shortens and the ticket stops being retyped. When it misleads, it misleads at production speed. A preset from the wrong file, a curve pushed for a different stock, a job released by a rule nobody is watching. The crew's defence is the same as it always was. Look at the sheet. The connection doesn't retire that. It does make it easier for the wrong instruction to look official, because it came through the screen rather than through a shout across the hall.
The workflow server is the plant's own memory. If it's healthy, a night lead can find the job. If it's dead, the question is whether the memory was copied anywhere a press can use, or whether the memory was the only copy. Files that live only on that server, presets that load only from it, a schedule that exists only as a page in a browser: you've built a plant that forgets everything when one box forgets. That's a fragile way to hold a Friday.
The vendor line is the one you control least. It's how a technician helps, and it's how a machine can ask a stranger's computer for permission to be a press. Used as a convenience, it's a gift on the night a fault has a name and your own people are stuck. Used as a condition of running, it's a bet that the line, the vendor's end, and their mood about your account will all be fine on the afternoon you're late. You can like the gift and refuse the bet. Plenty of plants have signed the bet without noticing, because the start-up sequence worked every day until the day it didn't.
Who is allowed to change a setting
Write this down as people, not as a diagram. The minder at the console may change ink, water and the local settings of the machine in front of them. That's the job. Prepress may send a preset or a file, and the minder may reject it when it doesn't match the plates or the stock. Planning may change the order of jobs, and the floor may mark that order when the room disagrees. A vendor technician may look, and may change something only while the crew can see the change and can say no. An automatic update may wait. It doesn't get to pick Friday afternoon because a calendar on someone else's side said so.
The failures here are dull, which is why they happen. A shared login for the planning screen, stuck to a monitor, used by whoever is nearest. When the planner leaves the company, the login still moves jobs, and you're never quite sure who moved them. A remote session that adjusts a curve while the minder is at the feeder, so the next pull is "wrong" and the argument starts in the ink. A prepress operator who re-sends an old coverage file over a new plate and goes home. None of these require a break-in. They require a permission that was broader than the duty.
Tighten the permissions to the duty, and make a change visible. The console should show that a preset arrived from outside, with a name and a time, and accepting it should be an act, not a background event. Remote support should begin when the plant asks, and end when the plant says the call is over. People who leave give their access back on the last day, including the shared passwords you pretended were temporary. You don't need a clever scheme to get this wrong or right. You need to be able to answer, on a Friday, who changed that setting, and to know the minder could have refused it. If you can't answer, the connection is already a risk, whether or not anyone outside has taken an interest.
A connected digital press adds one more version of the same question. Jobs can arrive from a customer portal or a vendor queue without passing a desk. That's a fine way to take ordinary work, and a poor way to print a file nobody in the building has released. Keep a release that a person does. The press can hold the file, rip it, wait. It shouldn't start because the outside queue said start. The person who releases is also the person who can say the colour aim is missing. A queue can't say that. It can only be eager.
What a dead link should be allowed to spoil
Decide, before the Friday, what the press is allowed to need. It is reasonable to need power, paper, plates or a local file, ink, and a crew. It is not reasonable to need a handshake with a server that cuts cartons, or a phone call to another country, in order to feed a sheet you've already accepted. If a licence check or a cloud ticket sits in the start path, ask what happens when that check can't be made. If the answer is "the press waits", you've agreed that a dead link outranks a customer. Change the agreement. A machine should degrade to a press, not to a locked door.
Presets won't load if the server is down. That's a nuisance, not a crisis, if the crew still knows how to set a key or call up a curve that is stored on the machine. It becomes a crisis if nobody has set a key by hand for a year and the only people who could are off. A dead link reveals whether the skill is still in the room. Keep the skill. Run a job from local settings on purpose, now and then, so Friday isn't the rehearsal. The same goes for the schedule. A printed list or a local copy of the day's work, taken while the server is alive, will carry you through an afternoon. A schedule that exists only in a browser will not.
Files are the sharp version. If the printable file is already at the press, a dead server is an inconvenience to planning and a mild one to the crew. If the printable file exists only on the server, the press is a feeder with nothing to feed. Get the file to the machine before you need it, especially on a Friday, especially on work you've promised. And if an update is waiting on the vendor line, "not this week" has to be a sentence the plant can say and the machine will obey. An update that restarts a press in the middle of a promised job is not support. It's an interruption you didn't schedule. Support can wait until the lorry has gone.
When the link dies, tell the truth quickly. The office server is down, so new tickets won't appear; the jobs already at the machines continue. Or: the vendor line is down, so remote help is unavailable; we are not waiting on it. What you don't say is "everything is stopped" unless you've built a plant where everything really does depend on the dead box. If you've built that, Friday is when you learn it, and Saturday is when you change it. The older press that kept running is a slightly embarrassing teacher. It ran because its needs were local. That's a specification, not nostalgia.
What you keep able to run alone
Make a short list of what must still work when the office is dark and the vendor can't hear you. Plates or files already at the machine. A way to set ink, curves and feeder by hand, or from settings stored on the press. Colour targets kept where the crew can open them without the cloud: on the console, in the job bag, both. A list of the day's jobs that doesn't vanish with the browser. A way to record what you did, even if that's a pencil, so Monday's system can be told the truth. A delivery note you can produce from something you control, if the customer needs paper to take the goods. Remote diagnostics, automatic presets and a live schedule are welcome back when the links return. They aren't the conditions for the first sheet.
You're allowed to be connected. The preset, the shared queue and the technician who can see the console are worth having. They're worth more if losing them costs you the help and not the shift. The risk isn't that a press sits on a network. The risk is that nobody decided which decisions may arrive from outside, and nobody checked that a dead line fails soft. Decide the people. Keep the local means. Let Friday be an ordinary annoyance, the kind a minder can swear at and then print through, rather than the day the handshake became the job.